1. Introduction

This article introduces the research background and importance of the growth pattern of primary words in spoken English. Single frequency words, which appear only once in a text, are the main component of low-frequency words and have a significant impact on vocabulary diversity and richness. Research has shown that high-frequency words account for a large proportion in English discourse, while the proportion of primary words is relatively high in different corpora, posing a challenge to natural language processing. This study aims to explore the growth curve and vocabulary ratio model of primary words in English speaking through computer-aided analysis of the BNC spoken language corpus, in order to fill the gap in existing research.

2. Literature review and research questions

This chapter explores the growth pattern of a single word in spoken English and analyzes language phenomena using power law. The power law is widely used in econometrics, such as Fan u0026 Altmann's (2007) study of compound word composition, and Zipf's (1935) proposal of Zipf's law to describe word frequency distribution. However, Zipf's law has limitations in describing the distribution of low-frequency words. Econometric linguistics also studies the interdependence and functional laws between language features. Carroll (1969) proposed a theoretical model for the relationship between vocabulary addition and text capacity. After the 1970s, the establishment of large corpora provided data support for the study of vocabulary growth models, such as Guiraud, Herdan, Hybrid, Tuldava, and Brunet models. Feng Zhiwei (2011) proposed the economic law of term formation and the law of decreasing vocabulary growth. Liu Haitao and Huang Wei (2012) emphasized the application of mathematical methods in language research. This study is based on corpus and power law, focusing on the changes in the ratio of primary words to vocabulary in spoken English, as well as the fit between power function models and observed data.

3. Research methods

This chapter introduces the research methods, including corpus sources, data processing, and model applications. The study is based on the BNC spoken language corpus, which contains 10 million words and is divided into two parts: population sampling and discourse type sampling. Vocabulary statistics are based on word entries and do not distinguish between uppercase and lowercase letters, excluding numbers, personal names, place names, and symbols. The text block size is set to 2250 words, with a total of 4450 text blocks, and the actual size is approximately 2100 words. The Tuldava model is used to analyze the relationship between vocabulary and text length. The model approximates a power function to express the relationship between the class symbol to form symbol ratio and text capacity. The parameters α, β, and γ are related to the text style, type, and capacity.

4. Research results

By analyzing the growth curve of English spoken vocabulary and the ratio curve of primary words to vocabulary, the trend of increasing vocabulary and primary words with the accumulation of morphological symbols in English spoken language was revealed. Research has found that the growth rate of vocabulary gradually slows down, while the increase in the number of words at once tends to be flat. The ratio curve of a single word to vocabulary shows a downward sloping L-shape, and as the accumulation of oral symbols increases, the ratio decreases. This is different from the research results of Fan (2010), which found that the ratio curve of primary words to vocabulary in written language shows a U-shaped growth trend. This difference may be related to the way spoken and written language is spoken. The growth pattern of the ratio curve between primary words and vocabulary in spoken language is significantly different from that in written language. The coverage of core vocabulary in spoken language increases with the length of the text, while primary words in written language contain more out of dictionary vocabulary. The reparameterized Tuldava model can better reflect the relationship between the ratio of primary words to vocabulary and text length, with a high degree of model fit. This model provides a new perspective for describing the frequency distribution of low-frequency words, which is helpful for extracting technical terms in spoken English.

5. Conclusion

The study on the growth pattern of primary words in spoken English reveals that the number of primary words in spoken English increases slowly with the accumulation of form symbols, and initially increases rapidly; The ratio of primary words to vocabulary shows a downward sloping "L model", with a rapid decrease when the number of symbols is between 2000 and 50000, followed by a gradual decrease thereafter; The reparameterized Tuldava model effectively describes the ratio of primary words to vocabulary. The study emphasizes the importance of mastering low-frequency words in professional fields and points out that corpus research has reference value for fields such as econometrics and natural language processing.