Entity identification model training
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an entity identification model. In one aspect, a method includes obtaining a plurality of complete sentences that each include entity text that references a first entity; for each complete sentence in the plurality of complete sentences: providing a first portion of the complete sentence as input to an entity identification model that determines a predicted entity for the first portion of the complete sentence, the first portion being less than all of the complete sentence; comparing the predicted entity to the first entity; and updating the entity identification model based on the comparison of the predicted entity to the first entity.
1. A system comprising:
one or more data processing apparatus; and
a data storage device storing instructions that, when executed by the one or more data processing apparatus, cause the one or more data processing apparatus to perform operations comprising:
obtaining a plurality of complete sentences that each include entity text that references a first entity;
for each complete sentence in the plurality of complete sentences:
providing a first portion of the complete sentence as input to an entity identification model that determines a predicted entity for the first portion of the complete sentence, the first portion being less than all of the complete sentence;
comparing the predicted entity to the first entity; and
updating the entity identification model based on the comparison of the predicted entity to the first entity.
2. The system of claim 1 , wherein the operations further comprise:
for each sentence for which the predicted entity does not match the first entity:
providing a second portion of the complete sentence as input to the entity identification model that determines a second predicted entity for the second portion of the complete sentence, the second portion being different from the first portion and being less than all of the complete sentence;
comparing the second predicted entity to the first entity; and
updating the entity identification model based on the comparison of the second predicted entity to the first entity.
3. The system of claim 2 , wherein updating the entity identification model based on the comparison of the predicted entity to the first entity comprises reducing a prediction confidence score for the predicted entity when the first portion of the complete sentence is provided as input to the entity identification model.
4. The system of claim 2 , wherein updating the entity identification model based on the comparison of the second predicted entity to the first entity comprises increasing a prediction confidence score for the second predicted entity when the second portion of the complete sentence is provided as input to the entity identification model.
5. The system of claim 1 , wherein the operations further comprise:
determining, for each of the plurality of complete sentences, that the entity text included in the complete sentence references the first entity based on the inclusion of the first entity in the complete sentence.
6. The system of claim 1 , wherein the first portion of the complete sentence includes a portion of the entity text.
7. The system of claim 6 , wherein the entity identification model determines one or more predicted entities for the portion of the entity text included in the first portion of the complete sentence and, for each of the one or more predicted entities, a prediction confidence score that indicates a likelihood that the predicted entity matches the first entity.