These three approaches to analyzing language span both word-order dependent (Coh-Metrix) and bag-of-word (LIWC, LDA) approaches. 

Word-order dependent approaches preserve the relationship between words in a text (sentence, paragraph, document, etc.) and can be understood in the context of syntax parse trees. Click below for example.

The 2011 article on Coh-Metrix provides a description of the features generated by this software. Coh-Metrix was developed at The University of Memphis in the Institute for Intelligent Systems by Art Graesser, Max Louwerse, and Danielle McNamara. There are five high-level aggregate features generated by a principal components analysis (PCA) – these include syntax simplicity, word concreteness, narrativity, deep cohesion, and shallow cohesion. Syntactic features have been used in many disciplines to understand the contextual aspects of language such as more concrete or abstract word choices and the use of more or less complex phrases.

LIWC (Linguistic Inquiry and Word Count) is a dictionary-based program developed at UT Austin by James Pennebaker. This program is a quick and powerful approach to analyzing language by counting the number of words in a text, primarily closed-class words such as pronouns, conjunctions, articles, and prepositions. 

LDA is an open-source process which can be analyzed in many ways, including R, Python, and many off-the-shelf programs. LDA assigns words into categories based on their proximity in the corpus, and these terms form “topics”. Researchers qualitatively assign names to the words in a topic, based on their relationship and expertise in the subject matter.