Automating the generation of lexical patterns for processing free text in clinical documents

Abstract

Objective: Many tasks in natural language processing utilize lexical pattern-matching techniques, including information extraction (IE), negation identification, and syntactic parsing. However, it is generally difficult to derive patterns that achieve acceptable levels of recall while also remaining highly precise.

Materials and methods: We present a multiple sequence alignment (MSA)-based technique that automatically generates patterns, thereby leveraging language usage to determine the context of words that influence a given target. MSAs capture the commonalities among word sequences and are able to reveal areas of linguistic stability and variation. In this way, MSAs provide a systemic approach to generating lexical patterns that are generalizable, which will both increase recall levels and maintain high levels of precision.

Results: The MSA-generated patterns exhibited consistent F1-, F.5-, and F2- scores compared to two baseline techniques for IE across four different tasks. Both baseline techniques performed well for some tasks and less well for others, but MSA was found to consistently perform at a high level for all four tasks.

Discussion: The performance of MSA on the four extraction tasks indicates the method's versatility. The results show that the MSA-based patterns are able to handle the extraction of individual data elements as well as relations between two concepts without the need for large amounts of manual intervention.

Conclusion: We presented an MSA-based framework for generating lexical patterns that showed consistently high levels of both performance and recall over four different extraction tasks when compared to baseline methods.

Keywords: information extraction; natural language processing; text mining.

Figure 1:

Determining the words that make up the context of a target is a nontrivial task. Using distance does not favor one context over the other for the example target.

Figure 2:

Pair-wise alignment of two sentences, where the light gray shaded boxes represent matched tokens (base tokens) while the dark gray shaded boxes represent tokens that have been inserted. White boxes indicate tokens that do not participate in the alignment.

Figure 3:

An example MSA generated from the common tokens of several different sentences that share the same target. The MSA clearly shows the areas of stability and variation among the different sentences.

Figure 4:

Matching a pattern with a sentence is a pair-wise alignment with the inclusion of wildcards that can be matched against any token. In addition, to be considered a successful match, all tokens within the pattern must correspond with a token in the.

Figure 5:

System workflow. Sentences containing targets are first identified from documents and a subsample is set aside for both generating and scoring patterns. Scoring is based on a manually annotated set of instances. The scored patterns are then matched against sentences containing yet-to-be-extracted targets.

Figure 6:

Extraction tasks. Task 1 extracts the relation between nodule types and anatomy locations, Task 2 extracts nodule sizes, Task 3 extracts the relation between sizes and anatomy locations, and Task 4 extracts the relation between sizes and dates. Some markings have been left out, to preserve clarity.