Interference word recognition method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application relate to a kind of interference word identification method, device, electronic equipment and storage medium, the method includes: obtaining to be identified text, the to be identified text is carried out word segmentation processing, obtain the word segmentation set of the to be identified text;For each word segmentation in the word segmentation set, obtain the target text after the to be identified text deletes the word segmentation;Determine the semantic difference degree between the to be identified text and the target text and user behavior difference degree;According to the user behavior difference degree and the semantic difference degree, identify whether the word segmentation is interference word.Thereby, interference word identification can be carried out from two aspects of semantic difference degree and user behavior difference degree, improve the accuracy of interference word identification of to be identified text.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of computer, and particularly relate to an interference word identification method and device, electronic equipment and storage medium. BACKGROUND

[0002] At present, in the video search service, the user query often contains some interference words which often appear and have little help for retrieval. Therefore, in the search process, due to the influence of interference words, retrieval results that do not meet the real needs of users are caused.

[0003] In the prior art, interference word identification is performed by term frequency / inverse document frequency, but since some words are interference words in a specific context scenario, they may also be very important words for retrieval in other scenarios, resulting in a very low accuracy of identifying interference words. SUMMARY

[0004] In view of this, in order to solve the above technical problems or at least partially solve the above technical problems, embodiments of the present application provide an interference word identification method and device, electronic equipment and storage medium.

[0005] In a first aspect, the embodiments of the present application provide an interference word identification method, which comprises:

[0006] Obtaining a to-be-identified text, performing word segmentation processing on the to-be-identified text to obtain a word segmentation set of the to-be-identified text;

[0007] For each word segmentation in the word segmentation set, obtaining a target text after deleting the word segmentation from the to-be-identified text;

[0008] Determining a semantic difference degree and a user behavior difference degree between the to-be-identified text and the target text;

[0009] According to the user behavior difference degree and the semantic difference degree, identifying whether the word segmentation is an interference word.

[0010] In a possible implementation, the determination of the semantic difference degree and the user behavior difference degree between the to-be-identified text and the target text comprises:

[0011] Calculating a first semantic value of the to-be-identified text and a second semantic value of the target text through a preset language model;

[0012] According to the first semantic value and the second semantic value, determining the semantic difference degree between the to-be-identified text and the target text;

[0013] Determining a first user behavior value corresponding to the to-be-identified text and a second user behavior value corresponding to the target text;

[0014] According to the first user behavior value and the second user behavior value, a user behavior difference degree between the to-be-identified text and the target text is determined.

[0015] In a possible implementation, the user behavior includes a search behavior; the determining of the first user behavior value corresponding to the to-be-identified text and the second user behavior value corresponding to the target text includes:

[0016] searching the to-be-identified text to determine a first search result quantity corresponding to the to-be-identified text;

[0017] searching the target text to determine a second search result quantity corresponding to the target text;

[0018] determining the first search result quantity as a first search behavior value, and determining the second search result quantity as a second search behavior value.

[0019] In a possible implementation, the user behavior further includes a click behavior; the determining of the first user behavior value corresponding to the to-be-identified text and the second user behavior value corresponding to the target text further includes:

[0020] searching the to-be-identified text to determine a first search result corresponding to the to-be-identified text;

[0021] determining a first click quantity of a user clicking the first search result;

[0022] searching the target text to determine a second search result corresponding to the target text;

[0023] determining a second click quantity of a user clicking the second search result;

[0024] determining the first click quantity as a first click behavior value, and determining the second click quantity as a second click behavior value.

[0025] In a possible implementation, the determining of the user behavior difference degree between the to-be-identified text and the target text according to the first user behavior value and the second user behavior value includes:

[0026] determining a search behavior difference degree between the to-be-identified text and the target text according to the first search behavior value and the second search behavior value;

[0027] determining a click behavior difference degree between the to-be-identified text and the target text according to the first click behavior value and the second click behavior value;

[0028] The search behavior difference degree and the click behavior difference degree are determined as a user behavior difference degree.

[0029] In a possible implementation, the identifying whether the word is a noise word according to the user behavior difference degree and the semantic difference degree comprises:

[0030] inputting the user behavior difference degree and the semantic difference degree into a trained noise word discrimination model to obtain a discrimination value corresponding to the word;

[0031] comparing the discrimination value with a preset discrimination threshold to identify whether the word is a noise word;

[0032] if the discrimination value is greater than or equal to the discrimination threshold, the word is identified as a non-noise word;

[0033] if the discrimination value is less than the discrimination threshold, the word is identified as a noise word.

[0034] In a possible implementation, before the obtaining of the target text after the word is deleted from the text to be identified, the method further comprises:

[0035] performing part-of-speech tagging on each word in the set of words to obtain part-of-speech tagging information of the word;

[0036] determining whether the part-of-speech tagging information is target part-of-speech tagging information;

[0037] if yes, the word is identified as a noise word;

[0038] if no, the step of obtaining the target text after the word is deleted from the text to be identified is performed.

[0039] In a possible implementation, the trained noise word discrimination model is obtained by the following method:

[0040] for each text in a preset text library, determining a noise word in the text, and labeling the noise word in the text;

[0041] performing word segmentation processing on the text labeled with the noise word to obtain a target set of words of the text labeled with the noise word;

[0042] for any target word in the target set of words, obtaining a remaining text after the target word is deleted from the text labeled with the noise word;

[0043] determining a target semantic difference degree and a target user behavior difference degree between the text labeled with the noise word and the remaining text;

[0044] According to the target semantic difference degree, the target user behavior difference degree, and the interference word annotation situation of the word segmentation, an interference word discrimination model is trained.

[0045] In a possible implementation, the determining of the target semantic difference degree and the target user behavior difference degree between the text annotated with the interference word and the remaining text comprises:

[0046] A third semantic value of the text annotated with the interference word and a fourth semantic value of the remaining text are calculated through a preset language model;

[0047] According to the third semantic value and the fourth semantic value, a target semantic difference degree between the text annotated with the interference word and the remaining text is determined.

[0048] A third user behavior value corresponding to the text annotated with the interference word and a fourth user behavior value corresponding to the remaining text are determined.

[0049] According to the third user behavior value and the fourth user behavior value, a target user behavior difference degree between the text annotated with the interference word and the remaining text is determined.

[0050] In a possible implementation, the user behavior comprises a search behavior; the determining of the third user behavior value corresponding to the text annotated with the interference word and the fourth user behavior value corresponding to the remaining text comprises:

[0051] The text annotated with the interference word is searched to determine a third search result quantity corresponding to the text;

[0052] The remaining text is searched to determine a fourth search result quantity corresponding to the remaining text;

[0053] The third search result quantity is determined as a third search behavior value, and the fourth search result quantity is determined as a fourth search behavior value.

[0054] In a possible implementation, the user behavior further comprises a click behavior; the determining of the third user behavior value corresponding to the text annotated with the interference word and the fourth user behavior value corresponding to the remaining text further comprises:

[0055] The text annotated with the interference word is searched to determine a third search result corresponding to the text;

[0056] A third click quantity of the user clicking the third search result is determined;

[0057] The remaining text is searched to determine a fourth search result corresponding to the remaining text;

[0058] determine a fourth click number of times that the user clicks the fourth search result;

[0059] determine the third click number of times as a third click behavior value, and determine the fourth click number of times as a fourth click behavior value.

[0060] In a possible implementation, the determining, according to the third user behavior value and the fourth user behavior value, of a target user behavior difference degree between the text marked by the interference word and the remaining text comprises:

[0061] determining, according to the third search behavior value and the fourth search behavior value, of a target search behavior difference degree between the text marked by the interference word and the remaining text;

[0062] determining, according to the third click behavior value and the fourth click behavior value, of a target click behavior difference degree between the text marked by the interference word and the remaining text;

[0063] determining the target search behavior difference degree and the target click behavior difference degree as a target user behavior difference degree.

[0064] In a second aspect, an embodiment of the present application provides an interference word recognition device, and the device comprises:

[0065] a set obtaining module, configured to obtain a to-be-recognized text, perform word segmentation processing on the to-be-recognized text, and obtain a word segmentation set of the to-be-recognized text;

[0066] a text obtaining module, configured to obtain, for each word segmentation in the word segmentation set, a target text obtained by deleting the word segmentation from the to-be-recognized text;

[0067] a difference degree determining module, configured to determine a semantic difference degree and a user behavior difference degree between the to-be-recognized text and the target text;

[0068] a recognition module, configured to recognize, according to the user behavior difference degree and the semantic difference degree, whether the word segmentation is an interference word.

[0069] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor and a memory, wherein the processor is configured to execute an interference word recognition program stored in the memory, so as to implement the interference word recognition method in any one of the first aspect.

[0070] In a fourth aspect, an embodiment of the present application provides a storage medium, wherein the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors, so as to implement the interference word recognition method in any one of the first aspect.

[0071] The technical scheme provided by the embodiment of the present application comprises the following steps: obtaining to-be-recognized text, performing word segmentation processing on the to-be-recognized text to obtain a word segmentation set of the to-be-recognized text; obtaining target text after deleting each word segmentation in the word segmentation set from the to-be-recognized text; determining a semantic difference degree and a user behavior difference degree between the to-be-recognized text and the target text; and identifying whether the word segmentation is an interference word according to the user behavior difference degree and the semantic difference degree. Thus, interference word identification can be performed from two aspects of the semantic difference degree and the user behavior difference degree, which improves the accuracy of interference word identification of the to-be-recognized text compared with interference word identification only through the semantic difference degree or through word frequency / inverse document frequency. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 An embodiment flowchart of the interference word identification method provided by the embodiment of the present application is shown in the following table.

[0073] Figure 2 An embodiment flowchart of another interference word identification method provided by the embodiment of the present application is shown in the following table.

[0074] Figure 3 An embodiment flowchart of the user behavior value determination method provided by the embodiment of the present application is shown in the following table.

[0075] Figure 4 An embodiment flowchart of another user behavior value determination method provided by the embodiment of the present application is shown in the following table.

[0076] Figure 5 An embodiment flowchart of the user behavior difference degree determination method provided by the embodiment of the present application is shown in the following table.

[0077] Figure 6 An embodiment block diagram of the interference word identification device provided by the embodiment of the present application is shown in the following table.

[0078] Figure 7 A structural schematic diagram of the electronic device provided by the embodiment of the present application is shown in the following table. DETAILED DESCRIPTION

[0079] In order to make the objects, technical schemes and advantages of the embodiments of the present application clearer, the technical schemes in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments but not all embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0080] The interference word recognition method provided by the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. The embodiments do not constitute a limitation on the embodiments of the present invention.

[0081] See Figure 1 The above is a flowchart of an embodiment of a method for identifying interference words provided by this invention. Figure 1 As shown, the process may include the following steps:

[0082] Step 101: Obtain the text to be recognized, perform word segmentation on the text to be recognized, and obtain the word segmentation set of the text to be recognized.

[0083] The text to be identified is used to identify interfering words in the query text. It can be the query text entered by the user, the video content text, or the audio content text; this embodiment of the invention does not impose any limitations on this. The word segmentation set includes at least one word segmentation.

[0084] In one embodiment, the text to be identified is a query text; the specific implementation of obtaining the text to be identified may include: outputting a visual interface, obtaining the query text input by the user through the visual interface, and determining that the query text is the text to be identified.

[0085] In another embodiment, the text to be identified is video content text; the specific implementation of obtaining the text to be identified may include: when a trigger operation on the currently playing video is detected, obtaining the target video frame image of the currently playing video; determining the video content text in the target video frame image, and determining that the video content text is the text to be identified.

[0086] The aforementioned triggering operation can be a click operation on the video screen (e.g., single click, double click, etc.), a long press operation, or a click operation on a specified icon in the video screen (e.g., single click, double click, etc.). This embodiment of the invention does not limit this.

[0087] For example, suppose the current video playback progress is 01:23:00, and suppose the video content text output by the currently playing video is: "Listening to the sound of water under the starlight, listening to the evening bell of the nearby village, listening to the sound of weary cattle grazing by the riverbank." Then, according to the above description, when a trigger operation on the currently playing video is detected, the target video frame image of the currently playing video is obtained, that is, the video frame image corresponding to the video progress of 01:23:00; the video content text in the target video frame image is determined, and this video content text is identified as the text to be recognized: "Listening to the sound of water under the starlight, listening to the evening bell of the nearby village, listening to the sound of weary cattle grazing by the riverbank."

[0088] In yet another embodiment, the text to be recognized is audio content text; and the specific implementation of obtaining the text to be recognized can include: when a triggering operation on the currently played audio is detected, obtaining the audio content text of the currently played audio, and determining that the audio content text is the text to be recognized.

[0089] For example, assuming that the audio content text output by the currently played audio is: "listen to the sound of water under the starlight, listen to the sound of the village bell at night, and listen to the sound of the tired ox eating grass by the riverbank". Then, according to the above description, when a triggering operation on the currently played audio is detected, the audio content text of the currently played audio is obtained, and it is determined that the audio content text is the text to be recognized: "listen to the sound of water under the starlight, listen to the sound of the village bell at night, and listen to the sound of the tired ox eating grass by the riverbank".

[0090] It should be noted that the above is only an example of the text to be recognized being query text, video content text, and audio content text. In actual applications, it can also be other forms of text, and the embodiments of the present application do not limit this.

[0091] In an embodiment, to accurately recognize the interference words in the text to be recognized, the text to be recognized can be subjected to word segmentation processing to determine a plurality of segmented words included in the text to be recognized. In this way, the segmented words included in the text to be recognized can be subjected to interference word recognition, thereby improving the recognition accuracy.

[0092] Optionally, the specific implementation of the above-mentioned word segmentation processing of the text to be recognized to obtain the segmented word set of the text to be recognized can include: using a preset word segmentation tool to perform word segmentation processing on the text to be recognized to obtain the segmented word set of the text to be recognized. The word segmentation tool can be a jieba word segmentation tool, an LTP word segmentation tool, etc., and the embodiments of the present application do not limit this. In this way, at least one segmented word included in the text to be recognized can be obtained.

[0093] For example, assuming that the text to be recognized is: "good mood every day". Then, according to the above description, the segmented word set of the text to be recognized obtained by using the preset word segmentation tool to perform word segmentation processing on the text to be recognized can be { "every day", "good mood"}.

[0094] Step 102: For each segmented word in the segmented word set, obtaining a target text after deleting the segmented word from the text to be recognized.

[0095] Step 103: Determining the semantic difference degree and the user behavior difference degree between the text to be recognized and the target text.

[0096] The following describes steps 102 and 103:

[0097] The target text is a text obtained by deleting any segmented word in the set of segmented words in the to-be-identified text. The semantic difference degree refers to the difference between the semantic integrity of the to-be-identified text and the semantic integrity of the target text, which can be a difference value or a change rate between the two, and the embodiments of the present application do not limit this.

[0098] The user behavior refers to the behavior of the user after searching the to-be-identified text / target text, which can be a search behavior on the text, a click behavior on the search result. For example, the number of searches using the text, or the search result after searching using the text, or the number of clicks on the search result, etc.

[0099] In practice, a language model-based word deletion difference method can be used to identify the interference word in the text, that is, by comparing the results of the semantic integrity scoring using the language model before and after word deletion, the influence of the segmented word on the user query text expression is evaluated, and then the interference word is identified. However, the ability of the language model to identify interference words by scoring semantic integrity is limited, which may result in inaccurate identification results. Therefore, in the embodiments of the present application, the user behavior of the user querying the to-be-identified text is obtained, and the user behavior difference degree between the to-be-identified text and the target text is determined, and the interference word is identified according to the semantic difference degree and the user behavior difference degree, thereby improving the accuracy of interference word identification.

[0100] Optionally, the to-be-identified text and the target text can be input into a preset language model to enable the language model to score according to the language integrity thereof to obtain the semantic scores corresponding to the two, and determine the semantic difference degree between the to-be-identified text and the target text according to the semantic scores of the two.

[0101] Optionally, the number of user behaviors of the user searching using the to-be-identified text and the number of user behaviors of the user searching using the target text can be counted, and the user behavior difference degree between the to-be-identified text and the target text can be determined according to the corresponding number of user behaviors of the two.

[0102] It should be noted that the above is only an example of determining the semantic difference degree and the user behavior difference degree between the to-be-identified text and the target text, and in actual application, the semantic difference degree and the user behavior difference degree can also be determined by other ways, and the embodiments of the present application do not limit this.

[0103] Step 104, identifying whether the segmented word is an interference word according to the user behavior difference degree and the semantic difference degree.

[0104] The interference word refers to a word that often appears in a query text but has little help for retrieval, for example, ah, ha ha, and the like.

[0105] In one embodiment, to identify whether a word segment is a distractor word, the aforementioned user behavior difference and semantic difference can be input into a trained distractor word discrimination model. Based on the discrimination result output by the distractor word discrimination model, the word segmentation is identified as a distractor word. In this way, distractor word identification can be achieved from both semantic difference and user behavior difference aspects, thereby improving the recognition accuracy.

[0106] This concludes the process. Figure 1 The process described is as follows.

[0107] pass Figure 1 As shown in the flowchart, in the technical solution of this invention, the text to be identified is obtained, and then segmented into words to obtain a set of segmented words. For each segmented word in the set, the target text after deleting the segmented word is obtained. The semantic difference and user behavior difference between the text to be identified and the target text are determined. Based on the user behavior difference and semantic difference, it is determined whether the segmented word is a distractor word. Therefore, distractor word identification can be performed from both semantic difference and user behavior difference perspectives. This improves the accuracy of distractor word identification in the text to be identified compared to identifying distractor words solely through semantic difference or through word frequency / inverse document frequency.

[0108] See Figure 2 This is a flowchart illustrating another embodiment of the interference word recognition method provided by this invention. Figure 2 As shown, the process may include the following steps:

[0109] Step 201: Obtain the text to be recognized, perform word segmentation on the text to be recognized, and obtain the word segmentation set of the text to be recognized.

[0110] For a detailed description of step 201, please refer to [link / reference]. Figure 1 The relevant description of step 101 will not be repeated here.

[0111] Step 202: For each word in the word segmentation set, perform part-of-speech tagging on the word segment to obtain the part-of-speech tagging information corresponding to the word segment.

[0112] Step 203: Determine whether the part-of-speech tagging information is the target part-of-speech tagging information. If yes, proceed to step 212; otherwise, proceed to step 204.

[0113] The following provides a unified description of steps 202 and 203:

[0114] The parts of speech mentioned above refer to the classification of words based on their characteristics. These parts of speech may include, but are not limited to: nouns, pronouns, verbs, adjectives, numerals, adverbs, prepositions, conjunctions, interjections, and onomatopoeia.

[0115] The above-mentioned word-tagging information includes at least one word segment in the word segmentation set and the corresponding relationship between word natures. For example: "This" is a pronoun, "2" is a numeral, etc.

[0116] In one embodiment, to improve the accuracy of identifying interfering words, the word nature of each word segment in the word segmentation set can be determined, and the word nature of the word segment can be tagged to obtain word-tagging information. Determine whether the word-tagging information is the target word-tagging information. If the word-tagging information is the target word-tagging information, it can be determined that the word segment corresponding to the word-tagging information is an interfering word; otherwise, if the word-tagging information is not the target word-tagging information, then subsequent operations can be performed to further identify whether the word segment is an interfering word.

[0117] For example, assume that the target word-tagging information is an interjection, and assume that the word segmentation set is {"sky", "really", "blue", "ah"}. Then, according to the above description, for each word segment in the word segmentation set, the word segment is tagged with its word nature, and the word-tagging information corresponding to the word segment is as shown in Table 1 below:

[0118] Table 1

[0119]

[0120]

[0121] As can be seen from Table 1 above, by determining whether the word-tagging information is the target word-tagging information, it can be determined that the word-tagging information of the word segment "ah" is the target word-tagging information, and it is identified that the word segment is an interfering word; the word-tagging information of the word segments "sky", "really" and "blue" is not the target word-tagging information, and subsequent operations can be performed.

[0122] Step 204: Obtain the target text after deleting the word segment from the text to be recognized.

[0123] Step​​​​​​​​​​​​​Optionally, the to-be-recognized text can be input into a preset language model to enable the language model to score according to the language integrity of the to-be-recognized text to obtain a first semantic value; the target text can be input into the language model to enable the language model to score according to the language integrity of the target text to obtain a second semantic value. Then, the semantic difference degree between the to-be-recognized text and the target text can be determined according to the first semantic value and the second semantic value.

[0128] Optionally, the first semantic value and the second semantic value can be substituted into a preset semantic difference degree calculation formula to obtain the semantic difference degree between the to-be-recognized text and the target text.

[0129] For example, assuming that the to-be-recognized text is "today is sunny", the to-be-recognized file is subjected to word segmentation processing to obtain a word segmentation set { "today", "is", "sunny"}, and it is assumed that the preset semantic difference degree calculation formula is a difference calculation formula. Then, according to the above description, for each word segmentation in the word segmentation set, the target text obtained by deleting the word segmentation from the to-be-recognized text is "today is". The to-be-recognized text is input into the preset language model to obtain a first semantic score of 100; the target text is input into the language model to obtain a second semantic score of 55, and the first semantic score and the second semantic score are substituted into the preset semantic difference degree calculation formula to obtain the semantic difference degree between the to-be-recognized text and the target text, which is 45.

[0130] For example, assuming that the to-be-recognized text is "today is sunny", the to-be-recognized file is subjected to word segmentation processing to obtain a word segmentation set { "today", "is", "sunny"}, and it is assumed that the preset semantic difference degree calculation formula is a difference calculation formula. Then, according to the above description, for each word segmentation in the word segmentation set, the target text obtained by deleting the word segmentation from the to-be-recognized text is "today is". The to-be-recognized text is input into the preset language model to obtain a first semantic score of 100; the target text is input into the language model to obtain a second semantic score of 55, and the first semantic score and the second semantic score are substituted into the preset semantic difference degree calculation formula to obtain the semantic difference degree between the to-be-recognized text and the target text, which is 45.

[0131] Through this processing, the semantic difference degree between the to-be-recognized text and the target text can be determined according to the first semantic value of the to-be-recognized text and the second semantic value of the target text.

[0132] Step 207, determining a first user behavior value corresponding to the to-be-recognized text and a second user behavior value corresponding to the target text.

[0133] Step 208, determining a user behavior difference degree between the to-be-recognized text and the target text according to the first user behavior value and the second user behavior value.

[0134] The following provides a unified description of steps 207 and 208:

[0135] As can be seen from the above description, the above user behavior can be a search behavior for text and a click behavior on search results. Correspondingly, the user behavior value can be a search behavior value and a click behavior value.

[0136] In one embodiment, the above user behavior may include a search behavior; then, for the specific implementation of determining the first user behavior value corresponding to the text to be recognized and the second user behavior value corresponding to the target text, reference can be made to Figure 3 As shown, it is a flowchart of an embodiment of a method for determining a user behavior value provided by an embodiment of the present invention. As Figure 3 As shown, this process may include the following steps:

[0137] Step 301: Search for the text to be recognized and determine the number of first search results corresponding to the text to be recognized.

[0138] Step 302: Search for the target text and determine the number of second search results corresponding to the target text.

[0139] Step 303: Determine the number of first search results as the first search behavior value and determine the number of second search results as the second search behavior value.

[0140] The following provides a unified description of steps 301 to 303:

[0141] Among them, the above number of search results refers to the number of search results output when searching for the text to be recognized. In the embodiment of the present invention, the user behavior includes a search behavior, and correspondingly, the user behavior value is a search behavior value.

[0142] In the embodiment of the present invention, the search behavior difference degree between the text to be recognized and the target text can be determined through the number of search results.

[0143] For example, assume that the text to be recognized is: "Have a good mood every day", the target text is: "Good mood", and assume that when searching for the text to be recognized, the number of first search results corresponding to the text to be recognized is determined to be 100, and when searching for the target text, the number of second search results corresponding to the target text is determined to be 10. Then, the number of first search results can be determined as the first search behavior value, and the number of second search results can be determined as the second search behavior value.

[0144] In addition, the search behavior difference degree between the text to be recognized and the target text can also be determined by counting the first search times using the text to be recognized for search and the second search times using the target text for search. The embodiment of the present invention does not limit this.

[0145] Optionally, the search times of the search using the file to be identified or the target text can be determined by searching the search records saved in the database.

[0146] For example, assuming that the text to be identified is "every day good mood", the target text is "good mood", and the first search time of the search using the text to be identified is 80 and the second search time of the search using the target text is 30, the first search time can be determined as the first search behavior value and the second search time can be determined as the second search behavior value.

[0147] Through the processing manner, the first search behavior value and the second search behavior value corresponding to the search behavior of the user can be determined.

[0148] In an embodiment, the user behavior can further include a click behavior; and the specific implementation of determining the first user behavior value corresponding to the text to be identified and the second user behavior value corresponding to the target text can refer to the embodiment of the method for determining the user behavior value provided by the embodiment of the application shown in Figure 4 Figure 4 The flowchart can include the following steps:

[0149] Step 401: search the text to be identified to determine the first search result corresponding to the text to be identified.

[0150] Step 402: determine the first click times of the user clicking the first search result.

[0151] Step 403: search the target text to determine the second search result corresponding to the target text.

[0152] Step 404: determine the second click times of the user clicking the second search result.

[0153] Step 405: determine the first click times as the first click behavior value and determine the second click times as the second click behavior value.

[0154] The steps 401 to 405 are described uniformly as follows:

[0155] As can be seen from the above description, the user behavior includes a click behavior, and accordingly, the user behavior value is a click behavior value. In the embodiment of the application, by counting the first click times of the search result of the search of the file to be identified and the second click times of the search result of the search of the target text, the first click times are determined as the first click behavior value and the second click times are determined as the second click behavior value.

[0156] ​For example, assume that the text to be recognized is: "Good mood every day", the target text is: "Good mood", and assume that the first click count of the first search result obtained after the user searches for the text to be recognized is 210, and the second click count of the second search result obtained after the user searches for the target text is 20. Then, it can be determined that the first click count is the first click behavior value, and the second click count is the second click behavior value.

[0157] Based on the above description, in one embodiment, for the specific implementation of determining the user behavior difference degree between the text to be recognized and the target text according to the first user behavior value and the second user behavior value, refer to Figure 5 As shown, it is a flowchart of an embodiment of a method for determining user behavior difference degree provided by an embodiment of the present invention. As Figure 5 shown, this process may include the following steps:

[0158] Step 501: Determine the search behavior difference degree between the text to be recognized and the target text according to the first search behavior value and the second search behavior value.

[0159] Step 502: Determine the click behavior difference degree between the text to be recognized and the target text according to the first click behavior value and the second click behavior value.

[0160] Step 503: Determine that the search behavior difference degree and the click behavior difference degree are the user behavior difference degree.

[0161] The following is a unified description of steps 501 to 503:

[0162] Among them, the above user behavior difference degree may be the user behavior difference value of the user searching for the text to be recognized and the target file, or the user behavior change rate of the user searching for the text to be recognized and the target file. The embodiments of the present invention do not limit this.

[0163] As can be seen from the above description, the user behavior includes the search behavior and the click behavior. Correspondingly, the user behavior value includes the search behavior value and the click behavior value. Therefore, the user behavior difference degree can be determined according to the search behavior difference degree and the click behavior difference degree.

[0164] Optionally, substitute the first search behavior value and the second search behavior value into a preset difference degree calculation formula (for example, the change rate calculation formula) to obtain the search behavior difference degree between the text to be recognized and the target text. Substitute the first click behavior value and the second click behavior value into the preset difference degree calculation formula to obtain the click behavior difference degree between the text to be recognized and the target text. In this way, the user behavior difference degree can be obtained.

[0165] Step 209, inputting the user behavior difference degree and the semantic difference degree into the trained interference word discrimination model to obtain a discrimination value corresponding to the segmented word.

[0166] Step 210, comparing the discrimination value with a preset discrimination threshold to identify whether the segmented word is an interference word; if the discrimination value is greater than or equal to the discrimination threshold, step 211 is performed; if the discrimination value is less than the discrimination threshold, step 212 is performed.

[0167] Step 211, identifying that the segmented word is not an interference word.

[0168] Step 212, identifying that the segmented word is an interference word.

[0169] The following uniformly describes steps 209 to 212:

[0170] The interference word discrimination model is used to discriminate whether the segmented word corresponding to the input object is an interference word, and mainly calculates the input object through an interference word discrimination calculation formula.

[0171] Based on the above description, in an embodiment, the trained interference word discrimination model can be obtained in the following manner: for each text in a preset text library, determine the interference word in the text, and label the interference word in the text. Then, the text after the interference word labeling is segmented to obtain a target segmented word set of the text after the interference word labeling. Then, for any target segmented word in the target segmented word set, the remaining text after deleting the target segmented word from the text after the interference word labeling is obtained, and the target semantic difference degree and the target user behavior difference degree between the text after the interference word labeling and the remaining text are determined, and then the interference word discrimination model is trained according to the target semantic difference degree, the target user behavior difference degree and the interference word labeling of the segmented word. In this way, the interference word discrimination model can be trained.

[0172] Optionally, the interference word in the text can be labeled by user behavior weak supervision, or the interference word in the text can be labeled by a preset interference word dictionary, and the present embodiment does not limit this.

[0173] As an optional implementation, the specific implementation of determining the target semantic difference and the target user behavior difference between the text marked with the interference word and the remaining text can include: calculating a third semantic value of the text marked with the interference word and a fourth semantic value of the remaining text through a preset language model, determining the target semantic difference between the text marked with the interference word and the remaining text according to the third semantic value and the fourth semantic value; determining a third user behavior value corresponding to the text marked with the interference word and a fourth user behavior value corresponding to the remaining text, and determining the target user behavior difference between the text marked with the interference word and the remaining text according to the third user behavior value and the fourth user behavior value.

[0174] Optionally, the user behavior can include a search behavior, and the specific implementation of determining the third user behavior value corresponding to the text marked with the interference word and the fourth user behavior value corresponding to the remaining text can include: searching the text marked with the interference word to determine a third search result quantity corresponding to the text; searching the remaining text to determine a fourth search result quantity corresponding to the remaining text; determining the third search result quantity as the third search behavior value, and determining the fourth search result quantity as the fourth search behavior value.

[0175] Optionally, the user behavior can also include a click behavior, and the specific implementation of determining the third user behavior value corresponding to the text marked with the interference word and the fourth user behavior value corresponding to the remaining text can include: searching the text marked with the interference word to determine a third search result corresponding to the text, and determining a third click quantity of the user clicking the third search result; searching the remaining text to determine a fourth search result corresponding to the remaining text, and determining a fourth click quantity of the user clicking the fourth search result; determining the third click quantity as the third click behavior value, and determining the fourth click quantity as the fourth click behavior value.

[0176] Based on the above description, the specific implementation of determining the target user behavior difference between the text marked with the interference word and the remaining text according to the third user behavior value and the fourth user behavior value can include: determining a target search behavior difference between the text marked with the interference word and the remaining text according to the third search behavior value and the fourth search behavior value; determining a target click behavior difference between the text marked with the interference word and the remaining text according to the third click behavior value and the fourth click behavior value; and determining the target search behavior difference and the target click behavior difference as the target user behavior difference.

[0177] In the embodiment of the present application, after the interference word discrimination model is trained, the user behavior difference degree and the semantic difference degree are input into the interference word discrimination model, and the discrimination value corresponding to the word segmentation can be obtained, and then the discrimination value is compared with the preset discrimination threshold value, if the discrimination value is greater than or equal to the discrimination threshold value, it is determined that the word segmentation corresponding to the discrimination value is not an interference word, otherwise, if the discrimination value is less than the discrimination threshold value, it is determined that the word segmentation corresponding to the discrimination value is an interference word. In this way, the interference word in the text to be recognized can be quickly and accurately recognized.

[0178] For example, it is assumed that the discrimination threshold value is 10, and it is assumed that the user behavior difference degree and the semantic difference degree are input into the interference word discrimination model, and the discrimination value corresponding to the word segmentation is 13. Then, according to the above description, the discrimination value is compared with the preset discrimination threshold value, it is determined that the discrimination value is greater than the discrimination threshold value, and therefore it is determined that the word segmentation corresponding to the discrimination value is not an interference word.

[0179] It is further assumed that the user behavior difference degree and the semantic difference degree are input into the interference word discrimination model, and the discrimination value corresponding to the word segmentation is 3. Then, according to the above description, the discrimination value is compared with the preset discrimination threshold value, it is determined that the discrimination value is less than the discrimination threshold value, and therefore it is determined that the word segmentation corresponding to the discrimination value is an interference word.

[0180] At this point, the process of Figure 2 related to the flowchart is completed.

[0181] Through this processing mode, interference word recognition can be performed from two aspects of semantic difference degree and user behavior difference degree, and the accuracy of interference word recognition of the text to be recognized can be improved.

[0182] Corresponding to the foregoing embodiment of the interference word recognition method, the present application also provides an embodiment block diagram of a device.

[0183] Referring to Figure 6 , an embodiment block diagram of an interference word recognition device provided by the embodiment of the present application is provided. As Figure 6 shown, the device comprises:

[0184] A set acquisition module 601 is configured to acquire a text to be recognized, perform word segmentation processing on the text to be recognized, and obtain a word segmentation set of the text to be recognized.

[0185] A text acquisition module 602 is configured to acquire, for each word segmentation in the word segmentation set, a target text obtained by deleting the word segmentation from the text to be recognized.

[0186] A difference degree determination module 603 is configured to determine a semantic difference degree and a user behavior difference degree between the text to be recognized and the target text.

[0187] The identification module 604 is configured to identify whether the segmented word is an interference word according to the user behavior difference degree and the semantic difference degree.

[0188] In a possible implementation, the difference degree determination module 603 includes (not shown in the figure):

[0189] The semantic value calculation sub-module is configured to calculate a first semantic value of the text to be identified and a second semantic value of the target text by using a preset language model.

[0190] The semantic difference degree determination sub-module is configured to determine a semantic difference degree between the text to be identified and the target text according to the first semantic value and the second semantic value.

[0191] The behavior value determination sub-module is configured to determine a first user behavior value corresponding to the text to be identified and a second user behavior value corresponding to the target text.

[0192] The behavior difference degree determination sub-module is configured to determine a user behavior difference degree between the text to be identified and the target text according to the first user behavior value and the second user behavior value.

[0193] In a possible implementation, the user behavior includes a search behavior, and the behavior value determination sub-module is specifically configured to:

[0194] search the text to be identified to determine a first search result quantity corresponding to the text to be identified;

[0195] search the target text to determine a second search result quantity corresponding to the target text;

[0196] determine the first search result quantity as a first search behavior value, and determine the second search result quantity as a second search behavior value.

[0197] In a possible implementation, the user behavior further includes a click behavior, and the behavior value determination sub-module is further configured to:

[0198] search the text to be identified to determine a first search result corresponding to the text to be identified;

[0199] determine a first click quantity of the first search result clicked by a user;

[0200] search the target text to determine a second search result corresponding to the target text;

[0201] determine a second click quantity of the second search result clicked by the user;

[0202] determine the first click number as a first click behavior value, and determine the second click number as a second click behavior value.

[0203] In a possible implementation, the behavior difference determination submodule is specifically configured to:

[0204] determine a search behavior difference between the to-be-recognized text and the target text according to the first search behavior value and the second search behavior value;

[0205] determine a click behavior difference between the to-be-recognized text and the target text according to the first click behavior value and the second click behavior value;

[0206] determine the search behavior difference and the click behavior difference as a user behavior difference.

[0207] In a possible implementation, the identification module 604 is specifically configured to:

[0208] input the user behavior difference and the semantic difference into a trained interference word discrimination model to obtain a discrimination value corresponding to the segmented word;

[0209] compare the discrimination value with a preset discrimination threshold to identify whether the segmented word is an interference word;

[0210] if the discrimination value is greater than or equal to the discrimination threshold, identify that the segmented word is not an interference word;

[0211] if the discrimination value is less than the discrimination threshold, identify that the segmented word is an interference word.

[0212] In a possible implementation, the apparatus further includes (not shown in the figure):

[0213] an information determination module configured to, before the step of obtaining the target text after deleting the segmented word from the to-be-recognized text, perform part-of-speech tagging on each segmented word in the segmented word set to obtain part-of-speech tagging information of the segmented word;

[0214] an information judgment module configured to judge whether the part-of-speech tagging information is target part-of-speech tagging information;

[0215] a segmented word identification module configured to, if yes, identify that the segmented word is an interference word;

[0216] a step execution module configured to, if no, execute the step of obtaining the target text after deleting the segmented word from the to-be-recognized text.

[0217] In a possible implementation, the apparatus further includes (not shown in the figure):

[0218] a model determining module, configured to obtain the trained interference word discrimination model in the following manner:

[0219] an interference word labeling submodule, configured to determine interference words in each text in a preset text library, and label the interference words in the text;

[0220] a set determining submodule, configured to perform word segmentation processing on the text with the labeled interference words to obtain a target word segmentation set of the text with the labeled interference words;

[0221] a text determining submodule, configured to obtain, for any target word segmentation in the target word segmentation set, remaining text of the text with the labeled interference words after the target word segmentation is deleted;

[0222] a difference degree determining submodule, configured to determine a target semantic difference degree and a target user behavior difference degree between the text with the labeled interference words and the remaining text;

[0223] a model training submodule, configured to train an interference word discrimination model according to the target semantic difference degree, the target user behavior difference degree, and the interference word labeling condition of the word segmentation.

[0224] In a possible implementation, the difference degree determining submodule includes (not shown in the figure):

[0225] a semantic value calculating unit, configured to calculate a third semantic value of the text with the labeled interference words and a fourth semantic value of the remaining text by using a preset language model;

[0226] a semantic difference degree determining unit, configured to determine a target semantic difference degree between the text with the labeled interference words and the remaining text according to the third semantic value and the fourth semantic value;

[0227] a behavior value determining unit, configured to determine a third user behavior value corresponding to the text with the labeled interference words and a fourth user behavior value corresponding to the remaining text;

[0228] a behavior difference degree determining unit, configured to determine a target user behavior difference degree between the text with the labeled interference words and the remaining text according to the third user behavior value and the fourth user behavior value.

[0229] In a possible implementation, the user behavior includes a search behavior; and the behavior value determining unit is specifically configured to:

[0230] perform a search on the text with the labeled interference words to determine a third search result quantity corresponding to the text;

[0231] searching the remaining text to determine a fourth number of search results corresponding to the remaining text;

[0232] determining the third number of search results as a third search behavior value, and determining the fourth number of search results as a fourth search behavior value.

[0233] In a possible implementation, the user behavior further includes a click behavior; and the behavior value determination unit is further configured to:

[0234] searching the text marked with the interference word to determine a third number of search results corresponding to the text marked with the interference word;

[0235] determining a third number of clicks of the third search results by the user;

[0236] searching the remaining text to determine a fourth number of search results corresponding to the remaining text;

[0237] determining a fourth number of clicks of the fourth search results by the user;

[0238] determining the third number of clicks as a third click behavior value, and determining the fourth number of clicks as a fourth click behavior value.

[0239] In a possible implementation, the behavior difference determination unit is specifically configured to:

[0240] determining a target search behavior difference between the text marked with the interference word and the remaining text according to the third search behavior value and the fourth search behavior value;

[0241] determining a target click behavior difference between the text marked with the interference word and the remaining text according to the third click behavior value and the fourth click behavior value;

[0242] determining the target search behavior difference and the target click behavior difference as a target user behavior difference.

[0243] Figure 7 a structural schematic diagram of an electronic device provided by an embodiment of the present application, Figure 7 The electronic device 700 shown includes at least one processor 701, a memory 702, at least one network interface 704 and a user interface 703. The various components in the electronic device 700 are coupled together through a bus system 705. It can be understood that the bus system 705 is used to realize the connection and communication between the components. In addition to including a data bus, the bus system 705 also includes a power bus, a control bus and a status signal bus. However, for the purpose of clear illustration, all kinds of buses are marked as the bus system 705 in Figure 7 the description.

[0244] The user interface 703 can include a display, a keyboard or click device (e.g., a mouse, a trackball), a touchpad or touchscreen, etc.

[0245] It can be understood that the memory 702 in the embodiments of the present application can be a volatile memory or a nonvolatile memory, or can include both volatile and nonvolatile memory. The nonvolatile memory can be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically EPROM (EEPROM), or a flash memory. The volatile memory can be a Random Access Memory (RAM) used as an external cache. By way of example, and not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 702 described herein is intended to include, without being limited to, these and any other suitable types of memory.

[0246] In some embodiments, the memory 702 stores the following elements, executable units or data structures, or a subset of them, or an extended set of them: an operating system 7021 and an application program 7022.

[0247] The operating system 7021 contains various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application program 7022 contains various application programs, such as a media player, a browser, etc., for implementing various application services. The program for implementing the method of the embodiments of the present application can be included in the application program 7022.

[0248] In the embodiments of the present application, the processor 701 is configured to execute the method steps provided by the embodiments by invoking the programs or instructions stored in the memory 702, specifically, the programs or instructions stored in the application program 7022. For example, the processor 701 is configured to execute the following steps:

[0249] obtaining to-be-recognized text, performing word segmentation processing on the to-be-recognized text to obtain a word segmentation set of the to-be-recognized text;

[0250] For each word segmentation in the word segmentation set, obtaining a target text after deleting the word segmentation from the to-be-recognized text;

[0251] determining a semantic difference degree and a user behavior difference degree between the to-be-recognized text and the target text;

[0252] According to the user behavior difference degree and the semantic difference degree, identifying whether the word segmentation is an interference word.

[0253] The method disclosed in the embodiments of the present application can be applied to the processor 701 or implemented by the processor 701. The processor 701 can be an integrated circuit chip having a signal processing capability. In the implementation process, each step of the above method can be completed by an integrated logic circuit or an instruction in the form of software in the processor 701. The processor 701 described above can be a general processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. Each method, step and logic block diagram disclosed in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software units in the code processor for execution. The software unit can be located in a random access memory, a flash memory, a read only memory, a programmable read only memory, an electrically erasable programmable memory, a register or other mature storage media in the art. The storage medium is located in the memory 702, and the processor 701 reads the information in the memory 702 and combines the hardware to complete the steps of the above method.

[0254] It can be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing units can be implemented within one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSP Devices), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, micro-controllers, microprocessors, other electronic units designed to perform the functions described herein, or a combination thereof.

[0255] For software implementation, the techniques described herein can be implemented with a processing unit that executes program components or modules. The software code can be stored in memory and executed by a processor. Memory can be implemented within the processor or external to the processor.

[0256] The electronic device provided by the embodiments can be an electronic device as shown in Figure 7 , can perform all steps of the interference word recognition method as described in Figure 1 and Figure 2 , and thus achieve the technical effects of the interference word recognition method as described in Figure 1 and Figure 2 . For details, please refer to the descriptions of Figure 1 and Figure 2 . For brevity, they will not be described here.

[0257] The embodiments of the present application also provide a storage medium (computer readable storage medium). The storage medium stores one or more programs. The storage medium can include a volatile memory, such as a random access memory; the storage medium can also include a non-volatile memory, such as a read-only memory, a flash memory, a hard disk, or a solid state disk; the storage medium can also include a combination of the above kinds of memories.

[0258] When the one or more programs stored in the storage medium can be executed by one or more processors to implement the interference word recognition method described above.

[0259] The processor is configured to execute the interference word recognition program stored in the memory to implement the following steps of the interference word recognition method executed on the side of the electronic device:

[0260] Obtaining a to-be-recognized text, performing word segmentation processing on the to-be-recognized text to obtain a word segmentation set of the to-be-recognized text;

[0261] For each word in the word set, obtain the target text after deleting the word from the text to be recognized;

[0262] Determine the semantic difference and user behavior difference between the text to be recognized and the target text;

[0263] According to the user behavior difference and the semantic difference, identify whether the word is a noise word.

[0264] Those skilled in the art should further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in the foregoing description in a general manner. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0265] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, software executed by a processor, or a combination of both. The software modules can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art.

[0266] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method of interfering word recognition, characterized by, The method comprises: acquiring a text to be identified, performing word segmentation processing on the text to be identified to obtain a word segmentation set of the text to be identified; for each word segmentation in the word segmentation set, acquiring a target text after the text to be identified is deleted from the word segmentation; determining a semantic difference degree and a user behavior difference degree between the text to be identified and the target text, wherein the semantic difference degree is determined according to a semantic score between the text to be identified and the target text, and the user behavior difference degree is determined according to a corresponding user behavior frequency between the text to be identified and the target text; according to the user behavior difference degree and the semantic difference degree, identifying whether the word segmentation is an interference word, comprising: inputting the user behavior difference degree and the semantic difference degree into a trained interference word discrimination model to obtain a discrimination value corresponding to the word segmentation; comparing the discrimination value with a preset discrimination threshold to identify whether the word segmentation is an interference word; if the discrimination value is greater than or equal to the discrimination threshold, identifying that the word segmentation is not an interference word; and if the discrimination value is less than the discrimination threshold, identifying that the word segmentation is an interference word.

2. The method of claim 1, wherein, The determination of the semantic difference degree and the user behavior difference degree between the text to be identified and the target text comprises: calculating a first semantic value of the text to be identified and a second semantic value of the target text through a preset language model; determining the semantic difference degree between the text to be identified and the target text according to the first semantic value and the second semantic value; determining a first user behavior value corresponding to the text to be identified and a second user behavior value corresponding to the target text; determining the user behavior difference degree between the text to be identified and the target text according to the first user behavior value and the second user behavior value.

3. The method of claim 2, wherein, The user behavior comprises a search behavior; the determination of the first user behavior value corresponding to the text to be identified and the second user behavior value corresponding to the target text comprises: searching the text to be identified to determine a first search result quantity corresponding to the text to be identified; searching the target text to determine a second search result quantity corresponding to the target text; determining the first search result quantity as a first search behavior value and determining the second search result quantity as a second search behavior value.

4. The method of claim 3, wherein, The user behavior further comprises a click behavior; the determination of the first user behavior value corresponding to the text to be identified and the second user behavior value corresponding to the target text further comprises: searching the text to be identified to determine a first search result corresponding to the text to be identified; determining a first click frequency of the user clicking the first search result; searching the target text to determine a second search result corresponding to the target text; determining a second click frequency of the user clicking the second search result; determining the first click frequency as a first click behavior value and determining the second click frequency as a second click behavior value.

5. The method of claim 4, wherein, The user behavior difference degree between the to-be-recognized text and the target text is determined according to the first user behavior value and the second user behavior value, and the user behavior difference degree is determined according to the first search behavior value and the second search behavior value. The user behavior difference degree between the to-be-recognized text and the target text is determined according to the first user behavior value and the second user behavior value, and the user behavior difference degree is determined according to the first search behavior value and the second search behavior value. The user behavior difference degree between the to-be-recognized text and the target text is determined according to the first user behavior value and the second user behavior value, and the user behavior difference degree is determined according to the first search behavior value and the second search behavior value. The search behavior difference degree and the click behavior difference degree are determined as the user behavior difference degree.

6. The method of claim 1, wherein, Before the to-be-recognized text is obtained by deleting the target text after the word segmentation, the following steps are further included: For each word in the word segmentation set, the word is tagged with a part of speech to obtain part of speech tagging information of the word; It is determined whether the part of speech tagging information is target part of speech tagging information; If yes, the word is identified as an interference word; If not, the step of obtaining the to-be-recognized text by deleting the target text after the word segmentation is performed.

7. The method of claim 1, wherein, The trained interference word discrimination model is obtained by the following method: For each text in a preset text library, determine the interference word in the text, and label the interference word in the text; The text with interference word labeling is subjected to word segmentation processing to obtain a target word segmentation set of the text with interference word labeling; For any target word in the target word segmentation set, the remaining text is obtained by deleting the target word from the text with interference word labeling; The target semantic difference degree and the target user behavior difference degree between the text with interference word labeling and the remaining text are determined; According to the target semantic difference degree, the target user behavior difference degree, and the interference word labeling condition of the word, an interference word discrimination model is trained.

8. The method of claim 7, wherein, The target semantic difference degree and the target user behavior difference degree between the text with interference word labeling and the remaining text are determined by the following method: The third semantic value of the text with interference word labeling and the fourth semantic value of the remaining text are calculated by a preset language model; According to the third semantic value and the fourth semantic value, the target semantic difference degree between the text with interference word labeling and the remaining text is determined; The third user behavior value corresponding to the text with interference word labeling and the fourth user behavior value corresponding to the remaining text are determined; According to the third user behavior value and the fourth user behavior value, the target user behavior difference degree between the text with interference word labeling and the remaining text is determined.

9. The method of claim 8, wherein, The user behavior includes search behavior; the third user behavior value corresponding to the text with interference word labeling and the fourth user behavior value corresponding to the remaining text are determined by the following method: The text with interference word labeling is searched to determine the third search result quantity corresponding to the text; The remaining text is searched to determine the fourth search result quantity corresponding to the remaining text; The third search result quantity is determined as a third search behavior value, and the fourth search result quantity is determined as a fourth search behavior value.

10. The method of claim 9, wherein, The user behavior further includes a click behavior; the determination of the third user behavior value corresponding to the text marked with the interference word and the fourth user behavior value corresponding to the remaining text further includes: searching the text marked with the interference word to determine a third search result corresponding to the text; determining a third click number of the third search result clicked by the user; searching the remaining text to determine a fourth search result corresponding to the remaining text; determining a fourth click number of the fourth search result clicked by the user; determining the third click number as a third click behavior value and the fourth click number as a fourth click behavior value.

11. The method of claim 10, wherein, The determination of the target user behavior difference between the text marked with the interference word and the remaining text according to the third user behavior value and the fourth user behavior value includes: determining a target search behavior difference between the text marked with the interference word and the remaining text according to the third search behavior value and the fourth search behavior value; determining a target click behavior difference between the text marked with the interference word and the remaining text according to the third click behavior value and the fourth click behavior value; determining the target search behavior difference and the target click behavior difference as a target user behavior difference.

12. An interfering word recognition apparatus characterized by comprising: The apparatus includes: a set obtaining module configured to obtain a to-be-recognized text, perform word segmentation processing on the to-be-recognized text, and obtain a word segmentation set of the to-be-recognized text; a text obtaining module configured to, for each word segmentation in the word segmentation set, obtain a target text obtained by deleting the word segmentation from the to-be-recognized text; a difference degree determining module configured to determine a semantic difference degree and a user behavior difference degree between the to-be-recognized text and the target text, wherein the semantic difference degree is determined according to a semantic score between the to-be-recognized text and the target text, and the user behavior difference degree is determined according to a corresponding user behavior number between the to-be-recognized text and the target text; a recognition module configured to recognize whether the word segmentation is an interference word according to the user behavior difference degree and the semantic difference degree, and specifically configured to: input the user behavior difference degree and the semantic difference degree into a trained interference word discrimination model to obtain a discrimination value corresponding to the word segmentation; compare the discrimination value with a preset discrimination threshold to recognize whether the word segmentation is an interference word; if the discrimination value is greater than or equal to the discrimination threshold, recognize that the word segmentation is not an interference word; and if the discrimination value is less than the discrimination threshold, recognize that the word segmentation is an interference word.

13. An electronic device, comprising: include: a processor and a memory, the processor being configured to execute an interference word recognition program stored in the memory to implement the interference word recognition method in any one of claims 1-11.

14. A storage medium, characterized by The storage medium stores one or more programs, which can be executed by one or more processors to implement the interference word recognition method in any one of claims 1-11.

Images