Learning Semantic Correspondences in Technical Documentation

Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1612–1622 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1148 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1612–1622 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1148 Learning Semantic Correspondences in Technical Documentation Kyle Richardson and Jonas Kuhn Institute of Natural Language Processing University of Stuttgart {kyle,jonas}@ims.uni-stuttgart.de Abstract We consider the problem of translating high-level textual descriptions to formal representations in technical documenta- tion as part of an effort to model the meaning of such documentation. We fo- cus specifically on the problem of learn- ing translational correspondences between text descriptions and grounded represen- tations in the target documentation, such as formal representation of functions or code templates. Our approach exploits the parallel nature of such documentation, or the tight coupling between high-level text and the low-level representations we aim to learn. Data is collected by min- ing technical documents for such paral- lel text-representation pairs, which we use to train a simple semantic parsing model. We report new baseline results on sixteen novel datasets, including the standard li- brary documentation for nine popular pro- gramming languages across seven natural languages, and a small collection of Unix utility manuals. 1 Introduction Technical documentation in the computer domain, such as source code documentation and other how- to manuals, provide high-level descriptions of how lower-level computer programs and utilities work. Often these descriptions are coupled with formal representations of these lower-level features, ex- pressed in the target programming languages. For example, Figure 1.1 shows the source code doc- umentation (in red/bold) for the max function in the Java programming language paired with the representation of this function in the underlying Java language (in black). This formal representa- tion captures the name of the function, the return 1. Java Documentation **Returns the greater of two long values * @param a an argument * @param b another argument * @return the larger of a and b * @see java.lang.Long#MAX VALUE */ public static long max(long a, long b) 2. Clojure Documentation (defn random-sample "Returns items from coll with random probability of prob (0.0 - 1.0)" ([prob coll] ...)) 3. PHP documentation (French) *Ajoute une valeur comme dernier ´el´ement * @param value La valeur ´a ajouter * @see ArrayIterations::next() */ public void append(mixed $value) Figure 1: Example source code documentation. value, the types of arguments the function takes, among other details related to the function’s place and visibility in the overall source code collection or API. Given the high-level nature of the textual anno- tations, modeling the meaning of any given de- scription is not an easy task, as it involves much more information than what is directly provided in the associated documentation. For example, cap- turing the meaning of the description the greater of might require having a background theory about quantity/numbers and relations between different quantities. A first step towards capturing the meaning, however, …

Content truncated to 500 words