According to Du (2013), methods of discourse information analysis (hereinafter “DIA” for short) can fall into four categories: independent analysis, dominant analysis, parallel analysis and subsidiary analysis.

1.      Independent Analysis

Independent analysis means that only DIA is employed to analyze a discourse. If other methods are used that can be integrated into the framework of DIA, this sort of analysis can also be attributed to independent analysis. When independent analysis is conducted, the researcher needs to sample an appropriate number of discourse information factors. Otherwise, independent analysis cannot give a complete portrait of the discourse. On the other hand, those selected discourse information factors must be representative enough to guarantee the reliability of the analysis.

2.      Dominant Analysis

As for dominant analysis, more than two analytical methods are made use of where DIA enjoys a paramount position. When doing this kind of analysis, the researcher usually analyzes the discourse information factors in a detailed way. Other analytical methods serve to back up DIA and play the minor part in the whole analysis. For example, in authorship attribution, discourse information factors and discourse style factors are listed as below.

 

Discourse Information Factors	Discourse Style Factors
Information knots Added value of information knots Number of information knots Tree-structure Combination of information knots Information sharing Information source Repetition of information Information flow Communicative function Information element	Lexicon Grammar Pronunciation Collocation Sentence length Word length Word frequency Spelling Mistake Upper and lower case Simplified Chinese character Ellipsis

In the table above, the underlined factors are adopted for authorship attribution. As most of the analyzed features belong to discourse information factors, this analysis can be regarded as a typical example of dominant analysis. In addition, every discourse information factor can be treated further. For instance, discourse information knots can be classified into 15 types and discourse information elements can also be divided into different functional sets.

3.      Parallel Analysis

Parallel analysis signifies that DIA and other analytical methods are equally important. One of the benefits lies in that DIA and other analytical methods can lend support to each other. In this way, nearly all the important features of a discourse are covered and an all-round analysis can be ensured.

But just as every coin has two sides, if three methods are used, conflicts among them will arise and difficulties of explaining the analytical outcome will multiply. If more than three methods are elected, probably the researcher cannot obtain a definite outcome. Moreover, if factors analyzed by all these methods overlap a lot, one-sidedness tends to be inevitable.

4.      Subsidiary Analysis

In subsidiary analysis, DIA occupies a subordinate position, which can prop up other methods. To illustrate this point, forensic speaker identification can be taken for example. Although discourse information cannot help define the physical characteristics of speech sounds, it can nevertheless predict the occurrence of a certain speech sound through analyzing discourse information types. What is more, we can discover the distribution of speech sounds by calculating the distribution of discourse information. When other methods cannot provide a satisfying analysis of speech sounds, DIA may offer some useful suggestions. Therefore, we cannot ignore the role of DIA in subsidiary analysis.