Lightweight tagging for disjoint entities
AssigneeInternational Business Machines Corporation
InventorsDiwakar Mahajan, Ananya Aniruddha Poddar, Bharath Dandala, Ching-Huei Tsou
Life of this patent
Estimating the term…
Abstract
Text data including at least named entities can be received. From the named entities, continuous entities, overlapping entities and disjoint entities can be identified. The overlapping entities can be transformed into continuous entities. The continuous entities, the transformed entities and the disjoint entities can be encoded. The encoded entities can be input to a machine learning language model to train the machine learning model to predict candidate entities. The predicted entities can be decoded to reconstruct the predicted entities.
Claims15
A computer-implemented method comprising:
- receiving text data including at least named entities;
- identifying from the named entities, continuous entities, overlapping entities and disjoint entities;
- transforming the overlapping entities to continuous entities;
- encoding the continuous entities, the transformed entities and the disjoint entities;
- inputting the encoded entities to a machine learning language model to predict candidate entities;
- and
- decoding the predicted entities to reconstruct the predicted entities,
- wherein the encoding uses a DBIO encoding scheme, wherein B and I tags are used to label tokens belonging to the continuous entities, DB and DI tags are used to label the disjoint entities, and O is used to label tokens outside of continuous entities and disjoint entities,
- wherein the decoding includes determining whether an entity has a DB tag, and responsive to determining that the entity has a DB tag, reconstructing the entity as a disjoint entity.
A system comprising:
- a hardware processor;
- a memory device coupled with the hardware processor;
- the hardware processor configured to:
- receive text data including at least named entities;
- identify from the named entities, continuous entities, overlapping entities and disjoint entities;
- transform the overlapping entities to continuous entities;
- encode the continuous entities, the transformed entities and the disjoint entities;
- input the encoded entities to a machine learning language model to predict candidate entities;
- and
- decode the predicted entities to reconstruct the predicted entities,
- wherein the hardware processors is configured to use a DBIO encoding scheme to encode, wherein B and I tags are used to label tokens belonging to the continuous entities, DB and DI tags are used to label the disjoint entities, and O is used to label tokens outside of continuous entities and disjoint entities, wherein the DBIO encoding scheme excludes H tags that label head entities.
A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a device to cause the device to:
- receive text data including at least named entities;
- identify from the named entities, continuous entities, overlapping entities and disjoint entities;
- transform the overlapping entities to continuous entities;
- encode the continuous entities, the transformed entities and the disjoint entities;
- input the encoded entities to a machine learning language model trained using encoded continuous entities, transformed entities and disjoint entities to predict candidate entities;
- and
- decode the predicted entities to reconstruct the predicted entities,
- wherein the device is caused to train the machine learning model using a DBIO encoding scheme, wherein B and I tags are used to label tokens belonging to the continuous entities, DB and DI tags are used to label the disjoint entities, and O is used to label tokens outside of continuous entities and disjoint entities,
- wherein to decode, the device is caused to determine whether an entity has a DB tag, and responsive to determining that the entity has a DB tag, reconstruct the entity as a disjoint entity.
Dependencies are inferred from the claim text.