List of text mining methods explained

Different text mining methods are used based on their suitability for a data set. Text mining is the process of extracting data from unstructured text and finding patterns or relations. Below is a list of text mining methodologies.

• Centroid-based Clustering: Unsupervised learning method. Clusters are determined based on data points.^[1]
  • Fast Global KMeans: Made to accelerate Global KMeans.^[2]
  • Global-K Means: Global K-means is an algorithm that begins with one cluster, and then divides in to multiple clusters based on the number required.
  • KMeans: An algorithm that requires two parameters 1. K (a number of clusters) 2. Set of data.
  • FW-KMeans: Used with vector space model. Uses the methodology of weight to decrease noise.
  • Two-Level-KMeans: Regular KMeans algorithm takes place first. Clusters are then selected for subdivision into subclasses if they do not reach the threshold.
  • Cluster Algorithm
    • • • Hierarchical Clustering Agglomerative Clustering: Bottom-up approach. Each cluster is small and then aggregates together to form larger clusters.^[3]
    • Divisive Clustering: Top-down approach. Large clusters are split in to smaller clusters.
  • Density-based Clustering: A structure is determined by the density of data points.^[4]
    • DBSCAN
  • Distribution-based Clustering: Clusters are formed based on mathematical methods from data.
    • • Expectation-maximization algorithm Collocation
• Stemming Algorithm
  • Truncating Methods: Removing the suffix or prefix of a word.
    • Lovins Stemmer: Removes longest suffix.
    • Porters Stemmer: Allows programmers to stem words based on their own criteria.
  • Statistical Methods: Statistical procedure is involved and typically results in affixes being removed.
    • N-Gram Stemmer: A set of 'n' characters that are consecutive taken from a word
    • Hidden Markov Model (HMM) Stemmer: Moves between states are based on probability functions.
    • Yet Another Suffix Stripper (YASS) Stemmer: Hierarchal approach in creating clusters. Clusters are then considered a set of elements in classes and their centroids are the stems.
  • Inflectional & Derivational Methods
    • Krovetz Stemmer: Changes words to word stems that are valid English words.
    • Xerox Stemmer: Removes prefixes.^[5]
• Term Frequency
  • Term Frequency Inverse Document Frequency
• Topic Modeling
  • Latent Semantic Analysis (LSA)
  • Latent Dirichlet Allocation (LDA)
  • Non-Negative Matrix Factorization (NMF)
  • Bidirectional Encoder Representations from Transformers (BERT)
• Wordscores: First estimates scores on word types based on a reference text. Then applies wordscores to a text that is not a reference text to get a document score. Lastly, documents that are not referenced are rescaled to then compare to the reference text.^[6]

Notes and References

1. Web site: 2018-01-15 . Different Types of Clustering Algorithm . 2024-04-04 . GeeksforGeeks . en-US.
2. Jalil . Abdennour Mohamed . Hafidi . Imad . Alami . Lamiae . Khouribga . Ensa . 2016 . Comparative Study of Clustering Algorithms in Text Mining Context . International Journal of Interactive Multimedia and Artificial Intelligence . en . 3 . 7 . 42 . 10.9781/ijimai.2016.376 . 1989-1660.
3. Web site: 2021-02-01 . Agglomerative Methods in Machine Learning . 2024-04-04 . GeeksforGeeks . en-US.
4. Web site: Hahsler . etal . Michael . dbscan: Fast Density-based Clustering with R . 4 March 2024 . cran.r-project.org.
5. Web site: Ganesh Jivani . Anjali . A Comparative Study of Stemming Algorithms .
6. Lowe . Will . 2008 . Understanding Wordscores . Methods and Data Institute, School of Politics and International Relations, University of Nottingham, Nottingham . 10.2139/ssrn.1095280 . 1556-5068.