Natural Language Programming Using Class Sequential Rules
12433Authors
nan
Abstract
This paper presents a system for Natural Language Programming using Class Se- quential Rules (CSR). The system recog- nizes a number of procedural primitives and operators. The domain of the system is presently limited to the world of num- bers and operations on numbers. We eval- uate the effectiveness of CSRs at the task of Natural Language Programming using an annotated corpus of programming in- structions in natural language, achieving a precision and recall of 85% and 64% re- spectively. We also compare the perfor- mance of a system trained on annotated data with that of a system using hand- crafted rules. 1
Keywords
Full Content
Proceedings of the 5th International Joint Conference on Natural Language Processing, pages 237–245, Chiang Mai, Thailand, November 8 – 13, 2011. c⃝2011 AFNLP Natural Language Programming Using Class Sequential Rules Cohan Sujay Carlos Aiaioo Labs Bangalore, India cohan@aiaioo.com Abstract This paper presents a system for Natural Language Programming using Class Se- quential Rules (CSR). The system recog- nizes a number of procedural primitives and operators. The domain of the system is presently limited to the world of num- bers and operations on numbers. We eval- uate the effectiveness of CSRs at the task of Natural Language Programming using an annotated corpus of programming in- structions in natural language, achieving a precision and recall of 85% and 64% re- spectively. We also compare the perfor- mance of a system trained on annotated data with that of a system using hand- crafted rules. 1 Introduction Since the early days of computing, there have been those who have longed for a programming sys- tem wherein the use of a formal symbolism for programming was not strictly necessary. Though high level languages greatly improve the ease of programming, it might be easier for a user to be allowed to communicate with an application in a natural language rather than a formal language with a specialized syntax. Moreover, a significant fraction of the population of the world is not con- versant in the English language or remains unfa- miliar with the Roman alphabet, and for them, a system for programming in their native tongues would be of great benefit. It is also easy to fore- see a future where speech recognition systems are so accurate and robust that users might seek to provide instructions by means of speech to their computers. Thus there seems to be a need for algorithms capable of accepting commands and programming instructions in unconstrained human languages. Type Arity Example if 2 or 3 If x is 2, say “Hi” unless 2 Exit unless x is 2 while 2 While x ≤ 2 . . . until 2 Till x is 2 add 1 to x continuation 1 Also, increment y assignment 2 Let x be 1 imperatives 0 to ∞ Display x questions 1 What is y? y/n questions 1 Is x equal to 2? Table 1: Types of Procedural Primitives. We present a Natural Language Programming system capable of accepting programming com- mands in a natural language, executing them and returning any requested results to the user. It is limited to the domain of real numbers and can be used to write programs to compute the values of various functions of real numbers, or to gener- ate different number series. The system can rec- ognize nine broad categories of procedural prim- itives, some of which are conditionals, loops, as- signments and function calls (imperatives). The complete list of types of procedural primitives with their arity and examples of usage can be found in Table 1. We also present and evaluate a novel approach for processing user instructions. Instead of …
Content truncated to 500 words
Labeled Text
This paper presents a system for Natural Language Programming using Class Sequential Rules (CSR). The system recognizes a number of procedural primitives and operators. The domain of the system is presently limited to the world of numbers and operations on numbers. We evaluate the effectiveness of CSRs at the task of Natural Language Programming using an annotated corpus of programming instructions in natural language, achieving a precision and recall of 85% and 64% respectively. We also compare the performance of a system trained on annotated data with that of a system using handcrafted rules. 1
Publication Details
- Paper ID:
- 12433
- Year:
- 2011
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan