Address identification method, device, computer equipment and storage medium

Abstract

The present invention proposes an address recognition method, device, computer equipment and storage medium, wherein the method includes: obtaining multiple address texts; grouping the multiple address texts according to the geographical location corresponding to each address text and / or the text type to which each address text belongs; identifying a first text part containing a code and a second text part other than the first text part for each address text; for at least two address texts in the same group, using a semantic similarity model to identify the second text part to obtain the semantic similarity of the second text part between the at least two address texts; identifying whether the at least two address texts belong to the same address according to the semantic similarity of the at least two address texts and / or the text matching of the first text part. Therefore, by grouping the address texts and performing address matching or address alignment on the address texts in the group, the accuracy of address matching or address alignment can be improved.

Description

Technical Field

[0001] The present invention relates to the field of computers, and in particular to an address recognition method, device, computer equipment and storage medium. Background Art

[0002] When different data sources, business sources or times generate address texts, the same address object may have multiple text representations. For example, the two address texts "Room 888, Building 8, Xingyue International, Majuqiao Town, Tongzhou District, Beijing" and "Room 888, Building 8, Xingyue International Community, Xingmao 3rd Street, Tongzhou District, Beijing" represent the same address. The process of identifying these two address texts as the same address and assigning corresponding address IDs is called address alignment or address matching. After address alignment, multiple data sources can be integrated to analyze the people, things and objects at the address.

[0003] Therefore, how to recognize different address texts as the same address is very important. Summary of the invention

[0004] The present invention aims to solve one of the technical problems in the related art at least to a certain extent.

[0005] The present invention proposes an address recognition method, apparatus, computer equipment and storage medium, so as to achieve address matching or address alignment by grouping address texts and performing address matching or address alignment on the address texts within the groups, thereby improving the accuracy of address matching or address alignment, that is, improving the accuracy of the same address recognition result.

[0006] The first embodiment of the present invention provides an address recognition method, comprising:

[0007] Get multiple address texts;

[0008] Grouping the multiple address texts according to the geographical locations corresponding to the respective address texts and / or the text types to which the respective address texts belong;

[0009] identifying, for each of the address texts, a first text portion including a code, and a second text portion other than the first text portion;

[0010] For at least two address texts in the same group, using a semantic similarity model to identify the second text portion to obtain the semantic similarity of the second text portion between the at least two address texts;

[0011] According to the semantic similarity of the at least two address texts and / or the text matching of the first text portion, it is identified whether the at least two address texts belong to the same address.

[0012] The address recognition method of the embodiment of the present invention obtains multiple address texts, and groups the multiple address texts according to the geographical location corresponding to each address text, and / or the text type to which each address text belongs, and recognizes the first text part containing the code and the second text part other than the first text part for each address text, and then, for at least two address texts in the same group, uses a semantic similarity model to recognize the second text part to obtain the semantic similarity of the second text part between at least two address texts, and finally, recognizes whether at least two address texts belong to the same address according to the semantic similarity of at least two address texts and / or the text matching of the first text part. Therefore, by grouping the address texts and performing address matching or address alignment on the address texts in the group, the accuracy of address matching or address alignment can be improved, that is, the accuracy of the same address recognition result can be improved.

[0013] A second embodiment of the present invention provides an address recognition device, comprising:

[0014] The acquisition module is used to obtain multiple address texts;

[0015] A grouping module, used for grouping the plurality of address texts according to the geographical location corresponding to each of the address texts and / or the text type to which each of the address texts belongs;

[0016] A segmentation module, used for identifying, for each of the address texts, a first text portion containing a code, and a second text portion other than the first text portion;

[0017] A similarity recognition module, for identifying the second text portion using a semantic similarity model for at least two address texts in the same group, so as to obtain the semantic similarity of the second text portion between the at least two address texts;

[0018] The address recognition module is used to recognize whether the at least two address texts belong to the same address based on the semantic similarity of the at least two address texts and / or the text matching of the first text part.

[0019] The address recognition device of the embodiment of the present invention obtains multiple address texts, and groups the multiple address texts according to the geographical location corresponding to each address text, and / or the text type to which each address text belongs, and identifies the first text part containing the code and the second text part other than the first text part for each address text, and then, for at least two address texts in the same group, uses a semantic similarity model to identify the second text part to obtain the semantic similarity of the second text part between at least two address texts, and finally, identifies whether at least two address texts belong to the same address according to the semantic similarity of at least two address texts and / or the text matching of the first text part. Therefore, by grouping the address texts and performing address matching or address alignment on the address texts in the group, the accuracy of address matching or address alignment can be improved, that is, the accuracy of the same address recognition result can be improved.

[0020] The third aspect of the present invention proposes a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the address identification method proposed in the first aspect of the present invention is implemented.

[0021] The fourth aspect of the present invention provides a non-temporary computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the address identification method provided in the first aspect of the present invention.

[0022] A fifth aspect of the present invention provides a computer program product, including a computer program as provided in the first aspect of the present invention, wherein the computer program implements the address identification method as provided in the first aspect of the present invention when executed by a processor.

[0023] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0025] Figure 1 A schematic diagram of a flow chart of an address identification method provided in Embodiment 1 of the present invention;

[0026] Figure 2 A schematic diagram of a flow chart of an address identification method provided in Embodiment 2 of the present invention;

[0027] Figure 3 A flowchart of an address identification method provided in Embodiment 3 of the present invention;

[0028] Figure 4 is a schematic diagram of a geographic grid in an embodiment of the present invention;

[0029] Figure 5 A flowchart of an address identification method provided in Embodiment 4 of the present invention;

[0030] Figure 6 A schematic diagram of the structure of an address recognition device provided in Embodiment 5 of the present invention;

[0031] Figure 7 A block diagram of an exemplary computer device suitable for implementing embodiments of the present invention is shown. DETAILED DESCRIPTION

[0032] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0033] The following describes the address recognition method, apparatus, computer device and storage medium according to the embodiments of the present invention with reference to the accompanying drawings.

[0034] Figure 1 The figure is a flowchart of the address identification method provided in the first embodiment of the present invention.

[0035] The embodiment of the present invention is described by taking the address recognition method configured in an address recognition device as an example. The address recognition device can be applied to any computer device so that the computer device can perform the address recognition function.

[0036] Among them, the computer device can be any device with computing capabilities, such as a PC (Personal Computer), a mobile terminal, etc. The mobile terminal can be, for example, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, and other hardware devices with various operating systems, touch screens and / or display screens.

[0037] like Figure 1 As shown, the address identification method may include the following steps:

[0038] Step 101, obtaining multiple address texts.

[0039] In the embodiment of the present invention, there is no limitation on the method for obtaining the address text. For example, the following methods may be used to exemplify the method for obtaining the address text.

[0040] As an example, multiple address texts can be obtained from multiple websites through web crawler technology.

[0041] It is understandable that government departments, e-commerce platforms, logistics platforms, food delivery platforms, financial platforms, transportation platforms, etc. will generate a large amount of address data. Therefore, as another example, multiple address texts can be obtained from multiple network platforms, multiple data sources, multiple business sources, etc.

[0042] It should be noted that the user may input multiple address texts, where the input method includes but is not limited to touch input (such as sliding, clicking, etc.), keyboard input, voice input, etc. Therefore, as another example, multiple address texts may be obtained in response to user operations. The user operation may be an input operation, a selection operation, or other user-triggered operation.

[0043] For example, multiple address texts may be input manually or by voice by the user. Alternatively, an OCR (Optical Character Recognition) may be performed on a picture selected by the user to obtain multiple address texts on the picture.

[0044] Furthermore, in order to make the formats of the acquired multiple address texts more standardized, after acquiring the multiple address texts, the multiple address texts may be preprocessed, wherein the preprocessing may include but is not limited to converting full-width characters to half-width characters, converting traditional Chinese characters to simplified Chinese characters, filtering special format characters, filtering meaningless characters (such as spaces, tabs, irrelevant punctuation marks, etc.), deleting postscripts, converting Roman numerals and Chinese numerals to Arabic numerals, etc.

[0045] Step 102: grouping a plurality of address texts according to the geographical location corresponding to each address text and / or the text type to which each address text belongs.

[0046] It should be noted that the source of the address text generated by the Internet often records the longitude and latitude of the address, where longitude and latitude are the most common features for describing the spatial location of the address, and the longitude and latitude obtained by GPS (Global Position System) and other methods are relatively accurate, and the position deviation generally does not exceed 1 kilometer. Therefore, in the embodiment of the present invention, the geographical location corresponding to the address text may include longitude and latitude information. In addition, in addition to longitude and latitude, the administrative region to which the address text belongs or the accompanying regional code (such as regional ID) can also describe the spatial scope of the address. Therefore, in the embodiment of the present invention, the geographical location corresponding to the address text may also include the administrative region or regional ID.

[0047] It is understandable that different address texts may have different spatial locations and / or be located in different areas. In order to improve the accuracy of the same address recognition results, in a possible implementation method of an embodiment of the present invention, each address text may be divided according to the geographical location, and the address texts belonging to the same area may be divided into the same group.

[0048] In an embodiment of the present invention, the above-mentioned method of grouping address texts according to geographic location can be called spatial grouping. Spatial grouping is to group address texts based on geographic location, and the address texts under the same spatial grouping may still contain complex forms, multiple granularity (i.e., address level) combinations and rich features. Therefore, in order to improve the accuracy of the recognition results, in another possible implementation method of the embodiment of the present invention, the address texts can be divided more finely according to the text type of the address texts, so that the address texts within a single group have a certain pattern of consistency.

[0049] The text type to which the address text belongs can be determined according to the place indicated by the address text, or can also be determined according to the address level contained in the address text.

[0050] Specifically, multiple address texts may be grouped according to the text type to which each address text belongs, wherein each address text belonging to the same group belongs to the same text type.

[0051] As an example, when the text type to which the address text belongs is determined based on the place indicated by the address text, for example, the address text can indicate places such as residence, business, enterprise, government, education, public place, etc., the address text indicating the same place can be regarded as the same text type, and the address texts of the same text type can be divided into the same group. For example, the address text indicating a residential place can be divided into group 1, the address text indicating a commercial place can be divided into group 2, the address text indicating an enterprise place can be divided into group 3, the address text indicating a government place can be divided into group 4, the address text indicating an educational place can be divided into group 5, and the address text indicating a public place can be divided into group 6.

[0052] As another example, when the text type to which the address text belongs is determined based on the address level contained in the address text, for example, the address text may contain address levels such as community, cell, building, floor, portal, etc., the address text containing the same address level may be treated as the same text type, and the address texts of the same text type may be divided into the same group. For example, the address text containing the address level of community may be divided into group A, the address text containing the address level of cell may be divided into group B, the address text containing the address level of building may be divided into group C, the address text containing the address level of floor may be divided into group D, and the address text containing the address level of portal may be divided into group E.

[0053] In another possible implementation of the embodiment of the present invention, multiple address texts may be grouped according to the geographical locations corresponding to the address texts and the text types to which the address texts belong. The address texts belonging to the same group are within the same area and have the same text type (for example, the places indicated by the address texts are the same, or the address levels contained in the address texts are the same).

[0054] Step 103, for each address text, identifying a first text portion including a code and a second text portion other than the first text portion.

[0055] It should be noted that most address texts may include road numbers, village group numbers, house numbers, building numbers, unit numbers, room numbers, etc., such as "Room 402, Unit 3, Building 2", "No. 31, Group 2, Village 1". The first text portion containing the code in the embodiment of the present invention refers to the text portion containing the above numbering, for example, the first text portion may be a text portion containing numbers and / or letters.

[0056] In a possible implementation of the embodiment of the present invention, for each address text, the address text may be intercepted from the first appearing letter or number to the last character in the address text to obtain a first text portion containing the code. After the first text portion is intercepted, the text portion of the address text other than the first text portion may be used as the second text portion.

[0057] For example, taking the address text of "Building 8, 888, Xingyue International Community, Majuqiao Town, Tongzhou District, Beijing", the first text may be "Building 8, 888", and the second text may be "Xingyue International Community, Majuqiao Town, Tongzhou District, Beijing".

[0058] It should be noted that in actual application, in addition to numbers or letters before words such as building, floor, unit, room, group, and number, letters or numbers may also exist before other words. For example, the address text is "Room 1, Building 413, R&F Shangyue Residence, Majuqiao Town, outside the Sixth Ring Road, Tongzhou District, Beijing". The Chinese numbers in the address text are converted into Arabic numerals, and the pre-processed address text is "R&F Shangyue Residence, Building 1, Building 413, Majuqiao Town, outside the Sixth Ring Road, Tongzhou District, Beijing". At this time, if the address text is directly truncated from the first appearance of the letter or number to the last character in the address text, the first text part obtained will be "R&F Shangyue Residence, Building 1, Building 413, Majuqiao Town, outside the Sixth Ring Road". Obviously, the truncation result is unreasonable.

[0059] Therefore, as another possible implementation of the embodiment of the present invention, the above-cut text fragments can also be screened in combination with screening rules such as grammar matching to obtain the first text part containing the code. For example, only the letters and numbers before and after the building number, building, building, unit, room, group, number, etc. in the above-mentioned text part can be retained.

[0060] It should be noted that the present invention only takes step 103 being executed after step 102 as an example. In actual application, in order to improve recognition efficiency, step 103 can also be executed in parallel with step 102, or step 103 can also be executed before step 102. The embodiment of the present invention does not limit this.

[0061] Step 104: for at least two address texts in the same group, a semantic similarity model is used to identify the second text portion, so as to obtain the semantic similarity of the second text portion between the at least two address texts.

[0062] In the embodiment of the present invention, the similarity between any two address texts can be measured by a semantic similarity model. Specifically, after the address texts are grouped, for each group, the semantic similarity model can be used to identify the second text part of any two address texts in the group to obtain the semantic similarity of the second text part between the two address texts.

[0063] Step 105: Identify whether the at least two address texts belong to the same address based on the semantic similarity of the at least two address texts and / or the text matching of the first text portion.

[0064] In the embodiment of the present invention, it is possible to identify whether the at least two address texts belong to the same address based on the semantic similarity of the at least two address texts and / or based on the text matching of the first text portion.

[0065] In a possible implementation of the embodiment of the present invention, for at least two address texts in the same group, whether the at least two address texts belong to the same address can be identified based on the semantic similarity of the at least two address texts. For example, it can be determined whether the semantic similarity of the at least two address texts is greater than a set threshold. If the semantic similarity of the at least two address texts is greater than the set threshold, the at least two address texts can be identified as the same address, and if the semantic similarity of the at least two address texts is less than or equal to the set threshold, the at least two address texts can be identified as different addresses.

[0066] It is understandable that when the first text parts of two address texts match, it cannot be recognized that the two address texts belong to the same address. For example, address text 1 is "Room 888, Building 8, Community A, Tongzhou District, Beijing", and address text 2 is "Room 888, Building 8, Community B, Tongzhou District, Beijing". The first text part of address text 1 is "Room 888, Building 8", and the first text part of address text 2 is "Room 888, Building 8". It can be seen that the first text parts of address text 1 and address text 2 match, but address text 1 and address text 2 are not the same address.

[0067] Therefore, in order to improve the accuracy of the recognition result, in another possible implementation method of the embodiment of the present invention, it can be determined whether the above-mentioned at least two address texts contain the first text part. When the above-mentioned at least two address texts contain the first text part, it can be determined whether the first text parts in the above-mentioned at least two address texts match. When the first text parts in the above-mentioned at least two address texts do not match, the above-mentioned at least two address texts can be identified as different addresses. When the first text parts in the above-mentioned at least two address texts match, it is further determined whether the semantic similarity of the above-mentioned at least two address texts is greater than a first threshold. When the first text parts in the above-mentioned at least two address texts match and the semantic similarity is greater than the first threshold, the above-mentioned at least two address texts can be identified as the same address. When the semantic similarity of the above-mentioned at least two address texts is less than or equal to the first threshold, the above-mentioned at least two address texts can be identified as different addresses.

[0068] In the case where the at least two address texts do not include the first text portion, it can be determined whether the semantic similarity of the at least two address texts is greater than a second threshold value. In the case where the semantic similarity of the at least two address texts is less than or equal to the second threshold value, the at least two address texts can be identified as different addresses. In the case where the semantic similarity of the at least two address texts is greater than the second threshold value, the at least two address texts can be identified as the same address. The second threshold value is greater than the first threshold value.

[0069] That is to say, in an embodiment of the present invention, the first text part can be used as an auxiliary basis for address matching. When the address text contains the first text part, a smaller threshold can be set, which is recorded as the first threshold in the embodiment of the present invention, to avoid missing the address text match due to the threshold being set too high. When the address text does not contain the first text part, a larger threshold can be set, which is recorded as the second threshold in the embodiment of the present invention, to improve the accuracy of the recognition result.

[0070] It should be noted that, since the feature complexity of the first coding part is relatively low, it can be standardized with high accuracy, that is, the first coding part can be converted into standard indexes such as "[index][village][index][group][index][number]", "[index][building][index][unit][index][room]", etc. For example, the standard indexes corresponding to "Room 888, Building 8" and "Room 888, Building 8" are both [index=8][building][index=808][room]. Therefore, in another possible implementation of the embodiment of the present invention, it can be determined whether the at least two address texts contain the first text part. In the case that the at least two address texts contain the first text part, it can be determined whether the standard indexes of the first text parts in the at least two address texts are the same. In the case that the standard indexes of the first text parts in the at least two address texts are different, the at least two address texts can be identified as different addresses. In the case that the standard indexes of the first text parts in the at least two address texts are the same, it is further determined whether the semantic similarity of the at least two address texts is greater than a first threshold. In the case that the standard indexes of the first text parts in the at least two address texts are the same and the semantic similarity is greater than the first threshold, the at least two address texts can be identified as the same address.

[0071] By using the above method, each address text in each group can be aligned to complete the alignment of all address texts. Optionally, in order to improve the efficiency of address alignment, each group can be aligned in parallel.

[0072] As an application scenario, a set of address texts can be given, and the address texts in the address text set can be aligned with each other through the above method, or one or more new address texts can be given, and the new address texts can be aligned with the address set in the address library and then added to the address library. After the addresses are aligned, multiple data sources can be integrated to analyze the people, things, and objects on the addresses.

[0073] The address recognition method of the embodiment of the present invention obtains multiple address texts, and groups the multiple address texts according to the geographical location corresponding to each address text, and / or the text type to which each address text belongs, and recognizes the first text part containing the code and the second text part other than the first text part for each address text, and then, for at least two address texts in the same group, uses a semantic similarity model to recognize the second text part to obtain the semantic similarity of the second text part between at least two address texts, and finally, recognizes whether at least two address texts belong to the same address according to the semantic similarity of at least two address texts and / or the text matching of the first text part. Therefore, by grouping the address texts and performing address matching or address alignment on the address texts in the group, the accuracy of address matching or address alignment can be improved, that is, the accuracy of the same address recognition result can be improved.

[0074] In a possible implementation of the embodiment of the present invention, the semantic similarity model in the above embodiment may include a general recognition model and a group recognition model, and the semantic similarity may be determined according to a first similarity output by the general recognition model and a second similarity output by the group recognition model. The above process is described in detail below in conjunction with Example 2.

[0075] Figure 2 This is a flow chart of the address identification method provided in the second embodiment of the present invention.

[0076] like Figure 2 As shown, the address identification method may include the following steps:

[0077] Step 201, obtaining multiple address texts.

[0078] Step 202: grouping multiple address texts according to the geographical location corresponding to each address text and / or the text type to which each address text belongs.

[0079] Step 203, for each address text, identifying a first text portion including a code and a second text portion other than the first text portion.

[0080] In the embodiment of the present invention, steps 201 to 203 may be implemented in any manner in the embodiments of the present invention, respectively. The embodiment of the present invention does not limit this and will not be described in detail.

[0081] Step 204: input the second text parts of at least two address texts in the same group into the universal recognition model to obtain an output first similarity.

[0082] In the embodiment of the present invention, the universal recognition model can adopt supervised training or unsupervised training, and the embodiment of the present invention does not limit the training method of the universal recognition model. The training corpus of the universal recognition model can come from multiple groups, and the second text part in the address text in the multiple groups can be used as the training corpus of the universal recognition model.

[0083] Specifically, the training corpus of the general recognition model can be the second text part of the address texts collected from multiple groups. The training corpus of multiple groups can constitute a large-scale corpus collection. Based on the corpus collection, a complex similarity measurement model with a large number of parameters can be trained, that is, the general recognition model. Through the training of a large amount of training corpus, the general recognition model can obtain better address semantic expression ability and generalization, and has a certain similarity measurement ability for the address texts of each group.

[0084] When training the universal recognition model, the similarity between any two training samples (i.e., the second text parts) in the training corpus can be annotated, and the universal recognition model can be trained using any two training samples with annotated similarity, so that the universal recognition model can learn the correspondence between any two second text parts and the similarity.

[0085] Therefore, in the embodiment of the present invention, the second text parts of at least two address texts may be input into the universal recognition model to obtain the output first similarity. For example, the first similarity is marked as S1.

[0086] Step 205: input the second text parts of the at least two address texts into the group recognition model corresponding to the group to which they belong, so as to obtain an output second similarity.

[0087] In the embodiment of the present invention, each group has a corresponding group recognition model, and the training corpus of the group recognition model can come from the corresponding group, and the second text part of the address text in the corresponding group can be used as the training corpus of the group recognition model. That is, in the present invention, each group recognition model is fitted to the address features in the second text part of the corresponding group to achieve a more accurate address matching effect.

[0088] The group recognition model may adopt supervised training or unsupervised training, and the embodiment of the present invention does not limit the training method of the group recognition model.

[0089] Among them, the group recognition model can adopt lightweight models such as bag of words and TF-IDF (Term Frequency-Inverse Document Frequency) to reduce the training cost and storage cost of the model and improve the training efficiency of the model.

[0090] It should be noted that the existing semantic similarity model represents words and sentences with similar semantics as similar feature vectors by learning the deep semantic embedding of words and sentences. However, unlike everyday language, address text contains a large number of proper nouns. If address matching is performed only based on semantics, the accuracy of the recognition result is low. If the language model learned from corpus data other than addresses is transferred to the address text recognition scenario, it is still impossible to accurately measure the similarity of proper nouns in the address text. For example, there are four adjacent communities in Nantong City, Jiangsu Province: "***Wushan New Garden**", "***Wushan Home**", "***Wushan Apartment**", and "***Wushan Garden**". The semantic similarity of these four communities is high, but the literals are different. In reality, these four communities are different address objects. At the same time, since the geographical locations of these four communities are relatively close, the four communities are the same at the coarse-grained address level (province, city, district, county, etc.). After grouping based on geographical location and / or text type, the text context of these four communities is also extremely similar. Only using a general recognition model to calculate the similarity and identify the same address based on the similarity is more likely to result in recognition errors.

[0091] Therefore, in the embodiment of the present invention, in order to improve the accuracy of the recognition result, the similarity of the literal level features of the address text can also be calculated through the group recognition model, which is recorded as the second similarity in the embodiment of the present invention. Specifically, the second text part of the at least two address texts can be input into the group recognition model corresponding to the group to which they belong to obtain the output second similarity. For example, the second similarity is marked as S2.

[0092] Step 206: Determine the semantic similarity of the second text portion between the at least two address texts based on the first similarity and the second similarity.

[0093] In the embodiment of the present invention, the semantic similarity can be determined according to the first similarity and the second similarity, wherein there is a positive relationship between the semantic similarity and the first similarity, and there is a positive relationship between the semantic similarity and the second similarity.

[0094] In a possible implementation of the embodiment of the present invention, the first similarity and the second similarity may be weighted according to a preset weight to obtain the semantic similarity of the second text portion between the at least two address texts. For example, if the semantic similarity is S and the preset weight is σ, then S may be:

[0095] S = σ*S1+(1-σ)*S2; (1)

[0096] Step 207: Identify whether the at least two address texts belong to the same address based on the semantic similarity of the at least two address texts and / or the text matching of the first text portion.

[0097] The execution process of step 207 may refer to the execution process of step 105 in the above embodiment, and will not be described in detail here.

[0098] The address recognition method of the embodiment of the present invention can improve the accuracy of the similarity settlement results by identifying the semantic similarity between different address texts through a general recognition model and a group recognition model, thereby improving the accuracy of subsequent recognition results of the same address.

[0099] In order to clearly illustrate how the present invention groups address texts according to geographic locations, this embodiment provides another address recognition method. Figure 3 This is a flow chart of the address identification method provided in Embodiment 3 of the present invention.

[0100] like Figure 3 As shown, the address identification method may include the following steps:

[0101] Step 301, obtaining multiple address texts.

[0102] In the embodiment of the present invention, step 301 may be implemented in any manner in the embodiments of the present invention, and the embodiment of the present invention does not limit this and will not be described in detail.

[0103] Step 302: Determine the area range to which each address text belongs from a plurality of set area ranges according to the geographical location corresponding to each address text.

[0104] Among them, each set area range is pre-set, and the size and shape of the set area range can be set according to actual application requirements. For example, the shape of the set area range can be a matrix, a triangle, a regular hexagon, etc., and the size of the set area range can be divided into units of villages, townships, towns, streets, districts, counties, etc., and the embodiments of the present invention do not limit this.

[0105] In the embodiment of the present invention, the area range to which each address text belongs can be determined from multiple set area ranges according to the geographical location corresponding to each address text. For example, when the geographical location corresponding to the address text is within a certain set area range, the area range to which the address text belongs can be determined to be the set area range.

[0106] In a possible implementation of an embodiment of the present invention, when the geographic location includes longitude and latitude information, the longitude and latitude information representing the geographic location may be hash-coded for each address text. For example, the longitude and latitude information representing the geographic location may be hash-coded using a GeoHash algorithm, and the region to which each address text belongs may be determined based on the hash code. Address texts with the same hash code belong to the same region. Therefore, in an embodiment of the present invention, address texts with the same hash code may be divided into the same group.

[0107] It should be noted that the above-mentioned grouping method based on the geographic hash algorithm is to divide the geographic space into completely mutually exclusive different areas, but in reality, the longitude and latitude collected by GPS and other methods may have certain deviations, and the address text at the edge of the hash coding grid may be inaccurately divided. Therefore, it is necessary to optimize the address text at the edge of the geographic grid based on the geographic grid grouping.

[0108] For example, participate Figure 4 , the address texts within the geographic grid and across the edge of the geographic grid can be divided into the same group, where Figure 4 All address texts within the range enclosed by the dotted circle in belong to the group centered on the hash code GeoHash4. In this way, the optimized adjacent groups can have overlapping address texts instead of being completely mutually exclusive.

[0109] Specifically, the hash codes belonging to the same geographic grid and the hash codes at the edge of the address network can be used to determine the area range corresponding to the geographic grid. The area range to which each address text belongs can be determined based on the area range where each hash code is located, so that the address texts belonging to the same area range can be divided into the same group.

[0110] Step 303: divide the address texts belonging to the same area range into the same group.

[0111] In the embodiment of the present invention, after the area range of each address text is determined, the address texts belonging to the same area range may be divided into the same group.

[0112] Step 304, identifying, for each address text, a first text portion including the code and a second text portion other than the first text portion.

[0113] Step 305: for at least two address texts in the same group, a semantic similarity model is used to identify the second text portion, so as to obtain the semantic similarity of the second text portion between the at least two address texts.

[0114] Step 306: Identify whether the at least two address texts belong to the same address based on the semantic similarity of the at least two address texts and / or the text matching of the first text portion.

[0115] In the embodiment of the present invention, steps 304 to 306 may be implemented in any manner in the embodiments of the present invention, respectively. The embodiment of the present invention does not limit this and will not be described in detail.

[0116] It should be noted that due to the grouping method of the above hash code, there may be overlapping address texts in adjacent groups. After address alignment or address matching is performed through the above steps, the alignment results of adjacent geographic grids can be fused to eliminate missed matches caused by longitude and latitude offsets, thereby improving the accuracy of recognition results.

[0117] Specifically, the repeated address texts belonging to at least two target groups can be queried, and the target address texts belonging to the same address as the repeated address texts can be determined in the at least two target groups, the repeated address texts can be deduplicated, and the target address texts in the at least two target groups can be determined to correspond to the same address.

[0118] For example, address text 1 is located at the edge of both address grid 1 and address grid 2. At this time, address text 1 exists in both group 1 and group 2. Assuming that address text 2 and address text 3 in group 1 belong to the same address as address text 1, and address text 4 and address text 5 in group 2 belong to the same address as address text 1, it can be determined that address text 1, address text 2, address text 3, address text 4 and address text 5 all belong to the same address. That is, the two address clusters aligned with address text 1 in adjacent groups can be merged.

[0119] It should be noted that the above-mentioned grouping process of address text is described by hash coding, but the address identification method in the embodiment of the present invention can be applied to a variety of grouping strategies. When the geographical location is an administrative area or an area ID, an appropriate grouping strategy can also be used to group the address texts. The address texts belonging to the same administrative area or the same area ID can be divided into the same group. The embodiment of the present invention does not impose any restrictions on this.

[0120] The address recognition method of the embodiment of the present invention can improve the efficiency of address alignment by grouping address groups. When aligning addresses, only the address texts within the same area, that is, within the same group, need to be aligned with each other, without matching the address texts in other groups that are far apart. In addition, there may be two or more address texts with the same text fragments in different areas, such as the same community name or road name. By aligning addresses within each group, it is helpful to reduce the error rate of address alignment or address matching and improve the accuracy of recognition results.

[0121] In order to clearly illustrate how the present invention groups address texts according to text types, this embodiment provides another address recognition method. Figure 5 This is a flow chart of the address identification method provided in Embodiment 4 of the present invention.

[0122] like Figure 5 As shown, the address identification method may include the following steps:

[0123] Step 401, obtaining multiple address texts.

[0124] In the embodiment of the present invention, step 401 may be implemented in any manner in the embodiments of the present invention, which is not limited in the embodiment of the present invention and will not be described in detail.

[0125] Step 402: Determine the text type to which the address text belongs according to the address level contained in each address text in the preset multi-level address.

[0126] In the embodiment of the present invention, the multi-level address is pre-set. For example, taking the multi-level address including 14 address levels as an example, the multi-level address can be as shown in Table 1.

[0127]

[0128]

[0129] Table 1 Address levels in a multi-level address

[0130] POI is the abbreviation of Points of Interest. POI refers to geographical entities closely related to the user's life, such as schools, banks, restaurants, gas stations, hospitals, supermarkets, a house, a shop, a mailbox, a bus stop, etc. In Table 1, level 1 is the largest address level, and level 14 is the smallest address level.

[0131] In an embodiment of the present invention, the text type to which the address text belongs can be determined according to the address level contained in each address text. For example, address texts with the same geographical level can be determined as the same text type. Wherein, when the address levels contained in the address text are different, the text types to which the address text belongs are different. For example, the text type corresponding to the geographical level of a community is different from the text type corresponding to the geographical level of a village group.

[0132] In a possible implementation manner of the embodiment of the present invention, the minimum address level included in each address text may be determined from the multi-level address, and the address texts with the same minimum address level may be regarded as the same text type.

[0133] Step 403: grouping the multiple address texts according to the text types to which the address texts belong.

[0134] In the embodiment of the present invention, after determining the text type to which each address text belongs, multiple address texts may be grouped according to the text type to which each address text belongs. For example, address texts of the same text type may be grouped into the same group.

[0135] In an embodiment of the present invention, the above-mentioned method of grouping address texts according to text type can be called semantic grouping. Optionally, when the minimum address levels corresponding to different address texts are the same, but the level types corresponding to the minimum address levels are different, they can also be divided into different groups. For example, the minimum address levels of address text A and address text B are both level 6, but the level type of address text A is village, and the level type of address text B is cell or community. At this time, address text A and address text B can be divided into different groups to improve the accuracy of address matching or address alignment results.

[0136] Therefore, by semantically grouping the address texts, the address texts in the group have the same minimum address level or level type or other feature pattern consistency, which can reduce the feature complexity during address matching.

[0137] Step 404, for each address text, identifying a first text portion including a code and a second text portion other than the first text portion.

[0138] Step 405: for at least two address texts in the same group, a semantic similarity model is used to identify the second text portion, so as to obtain the semantic similarity of the second text portion between the at least two address texts.

[0139] Step 406: Identify whether the at least two address texts belong to the same address based on the semantic similarity of the at least two address texts and / or the text matching of the first text portion.

[0140] In the embodiment of the present invention, steps 404 to 406 may be implemented in any manner in the embodiments of the present invention, respectively. The embodiment of the present disclosure does not limit this and will not be described in detail.

[0141] It should be noted that in the address matching or address alignment scenario of large-scale address texts, the massive address texts contain complex forms, multiple granularity combinations and rich features. Moreover, due to factors such as human errors or irregular filling in the process of generating address texts, the feature space of address texts is more complex. In the prior art, address texts of different forms, granularities and features are processed, calculated and judged using consistent types or rules, which limits the accuracy of address matching or address alignment.

[0142] In the embodiment of the present invention, each address text is strategically grouped so that the address texts in a single group have a certain pattern consistency, and the feature space of the address text is simplified. On this basis, a targeted semantic similarity recognition model (or rule) is used to perform detailed address matching, which can improve the accuracy of address matching. That is, in the present invention, since the general recognition model has learned good semantic expression ability through a large amount of training and achieved matching at the semantic level, the group recognition model adopts a lightweight model to fit the address texts in a single group, and achieves targeted address matching capabilities including literal similarity. The outputs of the two types of models are combined to obtain the semantic similarity between different address texts, which can improve the accuracy of the similarity calculation results. In addition, the first text part containing the code in the address text is used as a basis for verifying the address matching. With this basis, that is, when the address text contains the first text part, the similarity threshold can be set to be relatively loose, thereby reducing the situation where the address text is missed due to the threshold being set too high. Therefore, in the embodiment of the present invention, according to the semantic similarity between different addresses and the text matching of the first text part, it is identified whether they belong to the same address, and a more robust address matching effect can be achieved.

[0143] In addition, spatial grouping and / or semantic grouping of address texts can, on the one hand, achieve certain pattern consistency of address texts within a group, train a group recognition model using each address text within a group, and perform address matching based on the group recognition model, which can improve the accuracy of address matching results. On the other hand, multiple groups can perform address matching or address alignment in parallel, which can improve processing efficiency. Moreover, address matching only within each group can reduce the scope of address matching, which can also improve processing efficiency.

[0144] With the above Figures 1 to 5 Corresponding to the address recognition method provided in the embodiment, the present invention also provides an address recognition device. Figures 1 to 5 The address recognition method provided in the embodiment corresponds to the embodiment, so the implementation of the address recognition method is also applicable to the address recognition device provided in the embodiment of the present invention, and will not be described in detail in the embodiment of the present invention.

[0145] Figure 6 This is a schematic diagram of the structure of an address recognition device provided in Embodiment 5 of the present invention.

[0146] like Figure 6 As shown, the address identification device 600 may include: an acquisition module 601 , a grouping module 602 , a division module 603 , a similarity identification module 604 and an address identification module 605 .

[0147] The acquisition module 601 is used to acquire multiple address texts.

[0148] The grouping module 602 is used to group multiple address texts according to the geographical location corresponding to each address text and / or the text type to which each address text belongs.

[0149] The segmentation module 603 is used to identify, for each address text, a first text portion containing a code and a second text portion other than the first text portion.

[0150] The similarity identification module 604 is used to identify the second text part of at least two address texts in the same group using a semantic similarity model to obtain the semantic similarity of the second text part between the at least two address texts.

[0151] The address identification module 605 is used to identify whether at least two address texts belong to the same address based on the semantic similarity of at least two address texts and / or the text matching of the first text part.

[0152] Furthermore, in a possible implementation of the embodiment of the present invention, the address identification module 605 includes:

[0153] The first recognition unit is used to recognize at least two address texts as the same address if the first text parts in the at least two address texts match and the semantic similarity is greater than a first threshold when at least two address texts contain the first text part.

[0154] The second recognition unit is used to recognize at least two address texts as the same address if the semantic similarity is greater than a second threshold when at least two address texts do not contain the first text part; wherein the second threshold is greater than the first threshold.

[0155] Furthermore, in a possible implementation of the embodiment of the present invention, the semantic similarity recognition model includes a general recognition model and a group recognition model; the similarity recognition module 604 includes:

[0156] The first processing unit is used to input the second text parts of at least two address texts into the universal recognition model to obtain an output first similarity.

[0157] The second processing unit is used to input the second text parts of at least two address texts into the group recognition model corresponding to the group to which they belong, so as to obtain an output second similarity.

[0158] The determining unit is used to determine the semantic similarity according to the first similarity and the second similarity.

[0159] Furthermore, in a possible implementation manner of the embodiment of the present invention, the determination unit is specifically configured to: weight the first similarity and the second similarity according to a preset weight to obtain the semantic similarity.

[0160] Furthermore, in a possible implementation of the embodiment of the present invention, the grouping module 602 is specifically used to: determine the area range to which each address text belongs from multiple set area ranges according to the geographical location corresponding to each address text; and divide the address texts belonging to the same area range into the same group.

[0161] Furthermore, in a possible implementation of an embodiment of the present invention, the grouping module 602 is specifically used to: hash-code the longitude and latitude information representing the geographic location for each address text; determine the area range corresponding to the geographic grid using the hash codes belonging to the same geographic grid and the hash codes at the edge of the address network; and determine the area range to which each address text belongs based on the area range of each hash code.

[0162] Furthermore, in a possible implementation of the embodiment of the present invention, the address identification device 600 may also include:

[0163] The query module is used to query the repeated address texts belonging to at least two target groups, and to determine the target address texts belonging to the same address as the repeated address texts in the at least two target groups.

[0164] The deduplication module is used to deduplicate repeated address texts and determine the target address texts in at least two target groups as corresponding to the same address.

[0165] Furthermore, in a possible implementation of the embodiment of the present invention, the grouping module 602 is specifically used to: determine the text type to which the address text belongs according to the place indicated by the address text; and group multiple address texts according to the text type to which the address text belongs.

[0166] Furthermore, in a possible implementation of the embodiment of the present invention, the grouping module 602 is specifically used to: determine the text type to which the address text belongs according to the address level contained in each address text in the preset multi-level address; and group multiple address texts according to the text type to which the address text belongs.

[0167] Furthermore, in a possible implementation of the embodiment of the present invention, the grouping module 602 is specifically used to: determine the minimum address level contained in each address text from the multi-level address; and determine the address texts with the same minimum address level as the same text type.

[0168] The address recognition device of the embodiment of the present invention obtains multiple address texts, and groups the multiple address texts according to the geographical location corresponding to each address text, and / or the text type to which each address text belongs, and identifies the first text part containing the code and the second text part other than the first text part for each address text, and then, for at least two address texts in the same group, uses a semantic similarity model to identify the second text part to obtain the semantic similarity of the second text part between at least two address texts, and finally, identifies whether at least two address texts belong to the same address according to the semantic similarity of at least two address texts and / or the text matching of the first text part. Therefore, by grouping the address texts and performing address matching or address alignment on the address texts in the group, the accuracy of address matching or address alignment can be improved, that is, the accuracy of the same address recognition result can be improved.

[0169] In order to implement the above embodiments, the present invention also proposes a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the address identification method proposed in any of the above embodiments of the present invention is implemented.

[0170] In order to implement the above embodiments, the present invention further proposes a non-temporary computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the address identification method proposed in any of the above embodiments of the present invention is implemented.

[0171] In order to implement the above embodiments, the present invention further proposes a computer program product, including a computer program, which, when executed by a processor, implements the address identification method proposed in any of the above embodiments of the present invention.

[0172] Figure 7 A block diagram of an exemplary computer device suitable for implementing embodiments of the present invention is shown. Figure 7 The computer device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0173] like Figure 7 As shown, the computer device 12 is in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that connects various system components (including the system memory 28 and the processing unit 16).

[0174] The bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnection (PCI) bus.

[0175] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0176] The memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 7 not shown, usually called a "hard drive"). Although Figure 7Not shown in the figure, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a compact disc read only memory (CD-ROM), a digital versatile disc read only memory (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the various embodiments of the present invention.

[0177] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0178] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the computer device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the computer device 12 via a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0179] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the methods mentioned in the above embodiments.

[0180] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0181] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0182] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0183] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0184] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0185] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0186] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0187] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present invention. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. An address recognition method, characterized in that: The following steps are involved: Get multiple address texts; Grouping the multiple address texts according to the geographical locations corresponding to the respective address texts and / or the text types to which the respective address texts belong; identifying, for each of the address texts, a first text portion including a code, and a second text portion other than the first text portion; For at least two address texts in the same group, using a semantic similarity model to identify the second text portion to obtain semantic similarity of the second text portion between the at least two address texts; Identifying whether the at least two address texts belong to the same address based on the semantic similarity of the at least two address texts and the text matching of the first text portion; Wherein, the semantic similarity recognition model includes a general recognition model and a group recognition model; The step of using a semantic similarity model to identify the second text portion of at least two address texts in the same group to obtain the semantic similarity of the second text portion between the at least two address texts includes: Inputting the second text portion of the at least two address texts into the universal recognition model to obtain an output first similarity; Inputting the second text part of the at least two address texts into a group recognition model corresponding to the group to which it belongs, so as to obtain an output second similarity; The semantic similarity is determined according to the first similarity and the second similarity.

2. The address recognition method according to claim 1, characterized in that: The step of identifying whether the at least two address texts belong to the same address based on the semantic similarity of the second text portion of the at least two address texts and the text matching of the first text portion includes: In the case where the at least two address texts include the first text portion, if the first text portions in the at least two address texts match and the semantic similarity is greater than a first threshold, identifying the at least two address texts as the same address; In the case that the at least two address texts do not include the first text portion, if the semantic similarity is greater than a second threshold, the at least two address texts are identified as the same address; wherein the second threshold is greater than the first threshold.

3. The address recognition method according to claim 1, characterized in that: The determining the semantic similarity according to the first similarity and the second similarity includes: The first similarity and the second similarity are weighted according to a preset weight to obtain the semantic similarity.

4. The address recognition method according to any one of claims 1 to 3, characterized in that: The grouping of the plurality of address texts according to the geographical location corresponding to each of the address texts and / or the text type to which each of the address texts belongs includes: According to the geographical location corresponding to each of the address texts, determining the area range to which each of the address texts belongs from a plurality of set area ranges; The address texts belonging to the same area range are divided into the same group.

5. The address recognition method according to claim 4, characterized in that: The step of determining the area range to which each address text belongs from a plurality of set area ranges according to the geographical location corresponding to each address text includes: For each of the address texts, hash encoding is performed on the latitude and longitude information representing the geographical location; Determine the area range corresponding to the geographic grid according to the hash codes belonging to the same geographic grid and the hash codes at the edge of the geographic network; According to the area range where each of the hash codes is located, the area range to which each of the address texts belongs is determined.

6. The address recognition method according to claim 5, characterized in that: After identifying whether the at least two address texts belong to the same address based on the semantic similarity of the second text part of the at least two address texts and / or the text matching of the first text part, the method further includes: Searching for duplicate address texts belonging to at least two target groups, and determining target address texts belonging to the same address as the duplicate address text in the at least two target groups; The duplicate address texts are deduplicated, and the target address texts in the at least two target groups are determined to correspond to the same address.

7. The address recognition method according to any one of claims 1 to 3, characterized in that: The grouping of the plurality of address texts according to the geographical location corresponding to each of the address texts and / or the text type to which each of the address texts belongs includes: Determining the text type of the address text according to the place indicated by the address text; The multiple address texts are grouped according to the text types to which the address texts belong.

8. The address recognition method according to any one of claims 1 to 3, characterized in that: The grouping of the plurality of address texts according to the geographical location corresponding to each of the address texts and / or the text type to which each of the address texts belongs includes: Determining the text type to which the address text belongs according to the address level contained in each address text in the preset multi-level address; The multiple address texts are grouped according to the text types to which the address texts belong.

9. The address recognition method according to claim 8, characterized in that: The determining the text type to which the address text belongs according to the address level contained in each address text in the preset multi-level address includes: From the multi-level addresses, determining the minimum address level contained in each of the address texts; Address texts with the same minimum address level are defined as the same text type.

10. An address recognition device, characterized in that: include: The acquisition module is used to obtain multiple address texts; A grouping module, used for grouping the plurality of address texts according to the geographical location corresponding to each of the address texts and / or the text type to which each of the address texts belongs; A segmentation module, used for identifying, for each of the address texts, a first text portion containing a code, and a second text portion other than the first text portion; A similarity recognition module, for identifying the second text portion using a semantic similarity model for at least two address texts in the same group, so as to obtain the semantic similarity of the second text portion between the at least two address texts; An address recognition module, configured to recognize whether the at least two address texts belong to the same address according to the semantic similarity of the at least two address texts and the text matching of the first text portion; Wherein, the semantic similarity recognition model includes a general recognition model and a group recognition model; The similarity recognition module comprises: A first processing unit, configured to input the second text portion of the at least two address texts into the universal recognition model to obtain an output first similarity; A second processing unit, configured to input the second text portion of the at least two address texts into a group recognition model corresponding to the group to which it belongs, so as to obtain an output second similarity; A determining unit is used to determine the semantic similarity according to the first similarity and the second similarity.

11. The address recognition device according to claim 10, characterized in that: The address recognition module comprises: A first recognition unit is used for recognizing the at least two address texts as the same address if the first text parts in the at least two address texts match and the semantic similarity is greater than a first threshold value when the at least two address texts contain the first text part; The second recognition unit is used to recognize the at least two address texts as the same address when the at least two address texts do not contain the first text part and the semantic similarity is greater than a second threshold; wherein the second threshold is greater than the first threshold.

12. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the address recognition method as claimed in any one of claims 1 to 9 is implemented.

13. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the address recognition method as described in any one of claims 1 to 9 is implemented.

14. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the address recognition method according to any one of claims 1 to 9.

Images