Therefore, if you start python in the project’s root directory, the component’s fully qualified class name is addons.my_custom_components.EntityTypoFixer.
We will implement our custom component in a class named EntityTypoFixer in file named my_custom_components.py which is saved in a folder named addons relative to the project’s root directory. Solution: Building A Custom Component Step 1: Decide On The Custom Component’s Fully Qualified Class Name
#Typo synonym how to
The next section will describe how to build this solution. If the similarity between the extracted entity and an element in that list exceeds a certain threshold, we conclude that there’s probably a typo in the extracted entity and autocorrect it to that element. This isn’t ideal because not only will the list be unnecessarily long and harder to maintain, but we would have to anticipate possible mistakes our users can make!Ī better solution is to build a custom component that goes through the extracted entities and fuzzy matches them to a list of known entities. One potential solution is to include in the list of synonyms all possible typos (or at least common ones). This is really annoying because it is a common typo and failing to normalize it to some canonical form will interfere with downstream tasks that rely on the correct country entity being extracted. We see that that DIET Classifier manages to recognize united states of amercia as a country entity but since there is no exact match in the provided list of synonyms, it did not get map to usa unlike in the utterance in Figure 2. Figure 3: Extracting a country with a typo
0 kommentar(er)
