TL;DR
This commentary, written in 1970, reflects on the state of computational linguistics, lamenting the lack of discussion around the programming aspects of the field. The author proposes an informal session at a conference to address this gap, soliciting questions to stimulate discussion on programming choices, their influence on problem definition, and the practicalities of system development.
What Problem Does This Paper Address?
The author observes a disconnect in the computational linguistics community: while many researchers are developing specific computer implementations, there's a notable absence of discussion from the programmer's perspective. This gap is seen as regrettable because understanding the programming process – the choices made, the systems used, and the challenges encountered – is crucial for stimulating further innovation and sharing practical experiences. The commentary aims to bridge this gap by initiating a dialogue about the "how" of computational linguistics, not just the "what."
Methods
The "method" proposed is not a technical algorithm but a pedagogical and community-building approach. The author devises a set of questions designed to elicit detailed responses about the practicalities of computational linguistics projects. These questions cover:
- Project Type: Categorizing projects (IR, modeling, data processing, etc.) and their operational mode (one-time vs. continuous).
- Programming Environment: The specific language, hardware, and system used.
- Development Process: Whether batch processing, time-sharing, or conversational modes were employed, and the rationale behind these choices.
- Influence of Language: How the chosen programming language impacted problem definition and the programming itself.
- Resource Allocation: Time spent on thinking, physical work, and machine time, and how language choice affected these.
- System Design: Details on data structure definition, "system programming" (e.g., character manipulation), algorithm programming, debugging, and documentation.
- Practicalities: Program size, use of time-sharing, impact of "software bureaucracy" (like JCL), and the ease of transferring programs to different systems.
- Future Considerations: Whether the same language would be chosen again and experiences teaching programming to linguists.
The author intends to collect and summarize these responses to facilitate a more informed discussion at a conference.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- The "Missing Half" of Computational Linguistics: The core insight is that the field's progress is hindered by a lack of open discussion on the practical, often messy, realities of programming. This "half" includes the programmer's viewpoint, system design choices, and the influence of the development environment.
- Programming Choices Shape the Problem: The author posits that the selection of a programming language and system is not merely a technical detail but actively influences how a linguistic problem is defined and approached. This suggests a feedback loop where the tools shape the research questions.
- The Value of Informal Exchange: The commentary champions informal sessions and pre-conference correspondence as vital for sharing experiences that are too nuanced for formal papers. This highlights the importance of community and peer learning in a rapidly evolving technical field.
- Bridging the Gap Between Linguists and Programmers: The author's own project at KVAL, aiming to build a translation system for Braille, explicitly seeks to enable linguists with minimal programming training to contribute, suggesting a desire to democratize computational linguistics development.
Key Results
As this is a commentary and a call for discussion, there are no quantitative results presented. The "key result" is the articulation of the problem and the proposed method for addressing it: a structured inquiry to gather practical programming experiences from the computational linguistics community.
Limitations and Future Work
The author readily acknowledges the "naivety" of the questions, attributing it partly to their own inexperience in the field and partly to a deliberate design choice to avoid leading the respondents. The primary limitation is the reliance on voluntary participation and the potential for insufficient interest to warrant a session.
Future work is explicitly outlined: the author intends to edit and summarize the collected replies to make them available to participants before a conference, thereby fostering a more informed and productive discussion. The ultimate goal is to encourage a deeper understanding and sharing of practical programming knowledge within computational linguistics.
Summary updated