Learning Latent Structures for Cross Action Phrase Relations in Wet Lab Protocols

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 6737–6750 August 1–6, 2021. ©2021 Association for Computational Linguistics 6737 Learning Latent Structures for Cross Action Phrase Relations in Wet Lab Protocols Chaitanya Kulkarni, Jany Chan, Eric Fosler-Lussier, Raghu Machiraju Department of Computer Science and Engineering Ohio State University {kulkarni.132,chan.206,fosler-lussier.1,machiraju.1}@osu.edu Abstract Wet laboratory protocols (WLPs) are critical for conveying reproducible procedures in bi- ological research. They are composed of in- structions written in natural language describ- ing the step-wise processing of materials by specific actions. This process flow descrip- tion for reagents and materials synthesis in WLPs can be captured by material state trans- fer graphs (MSTGs), which encode global tem- poral and causal relationships between actions. Here, we propose methods to automatically generate a MSTG for a given protocol by ex- tracting all action relationships across multiple sentences. We also note that previous corpora and methods focused primarily on local intra- sentence relationships between actions and en- tities and did not address two critical issues: (i) resolution of implicit arguments and (ii) es- tablishing long-range dependencies across sen- tences. We propose a new model that incre- mentally learns latent structures and is better suited to resolving inter-sentence relations and implicit arguments. This model draws upon a new corpus WLP-MSTG which was created by extending annotations in the WLP corpora for inter-sentence relations and implicit argu- ments. Our model achieves an F1 score of 54.53% for temporal and causal relations in protocols from our corpus, which is a signifi- cant improvement over previous models - Dy- GIE++:28.17%; spERT:27.81%. We make our annotated WLP-MSTG corpus available to the research community. 1 1 Introduction Wet laboratory protocols (WLPs) play an integral role in bioscience and biomedical research by serv- ing as a vehicle to communicate experimental in- structions that allow for standardization and repli- cation of experiments. These procedures, typically written in natural language, prescribe actions (Fig- ure 1) to be conducted on materials that generally 1The dataset and code is available on the authors’ websites Isolation of temperate phages by plaque agar overlay 1. Grow the bacteria overnight. 2. ... 3. ... 4. Remove one tube of soft agar from the water bath. 5. Add 1.0 mL host culture and either 1.0 or 0.1 mL viral concentrate. Figure 1: Extraction of a MSTG from an example WLP. The MSTG is composed of Action Graphs (in grey), connected by temporal and causal relationships (e.g., temporal relation ”Site” between Remove and Add). The arrow indicate the direction of material flow. produce new materials which, in turn, are used by future actions to make newer materials. However, WLPs can be unclear, composed of disconnected and distant parts, and built upon implicit informa- tion that were referenced earlier or omitted entirely. Lack of careful documentation has led to a repro- ducibility crisis (Baker, 2016) in the biosciences and also poses considerable challenges for automa- tion of laboratory procedures: gleaning the effect and …

Content truncated to 500 words