Text intention recognition method and device and storage medium
A recognition method and text technology, applied in text database query, text database clustering/classification, unstructured text data retrieval, etc., can solve the problem of wrong intention, error, low sentence intention recognition rate, etc.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
example 1
[0138] Example 1: For the situation that the entity cannot be matched in the entity dictionary library.
[0139] Exemplarily, the text to be recognized is matched with each entity in the entity dictionary database, but no corresponding entity is matched, that is, the type of the entity in the text to be recognized cannot be determined. At this time, there is no process of fusing the second feature vector of the entity with the feature vector of the word, and the network model will determine the intention of the text to be recognized according to the semantics of the text to be recognized. Because, in the process of training the network model, for the case where the type of the entity is uncertain, training samples with various intentions are constructed. Therefore, the network model will treat the situation in Example 1 as a multi-intent scene, and the output candidate intent is a general intent. Then, in the process of post-processing the common intent, all intents correspon...
example 2
[0141] Example 2: For the case where all the matched entities in the entity dictionary are wrong.
[0142] Optionally, one or more wrong entities are matched. In this application, one wrong entity is matched as an example for illustration. The situation of matching multiple wrong entities is similar to that of matching one wrong entity, and will not be described again.
[0143] Exemplarily, the text to be recognized is matched with each entity in the entity dictionary database. Since the entities in the entity dictionary database are relatively noisy, wrong entities may be matched during the matching process. For example, the text to be recognized is "Please play me the old days", but there may only be the video entity "Old Times" in the entity dictionary. Therefore, during the matching process, the matched entity is the video entity "Old Time", that is, the entity matching result of the text to be recognized is "[CLS] Please play Hello Old Time for me [SEP][video name]7 :9",...
example 3
[0145] Example 3: For the entities matched in the entity dictionary, there are both correct and incorrect cases.
[0146] Optionally, the number of matched correct entities and incorrect entities can be one or more. In this application, a wrong entity and a correct entity are matched as an example for illustration, and other numbers of correct entities and wrong entities and this type will not be described again.
[0147] Exemplarily, the text to be recognized is matched with each entity in the entity dictionary database. Since the entities in the entity dictionary database are relatively noisy, multiple entities will be matched during the matching process. There may be both The right entity, and the wrong entity. For example, the text to be recognized is "Please play me the good old time", but there may be video entity "Hello old time" and audio entity "Old time" in the entity dictionary. Therefore, during the matching process, the matched entities include music entities an...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More