The LDATrainer function outputs:
- Message
- Model table
- Output table [Optional]
| Column Name | Data Type | Description |
|---|---|---|
| message | TEXT, VARCHAR, or VARCHAR(n) | Reports the iteration steps and perplexity of the model. The perplexity formula is: perplexity = 2 H (p) = 2-Σ x p (x) log2 p (x) where H (p) is the entropy of the distribution. There is no uniform standard for using perplexity to determine whether a model is good—perplexity varies with the training documents. However, you can use perplexity to find the best model for a specified set of training documents: Generate models for several subsets of the training documents and then choose the model with the lowest perplexity. |
| Column Name | Data Type | Description |
|---|---|---|
| topicid | INTEGER | Internally generated topic identifier. |
| value | BYTEA | Model in binary format. |
| Column Name | Data Type | Description |
|---|---|---|
| docid | Same as data type of doc_column in input table | Contains document identifiers from the input table. |
| topicid | INTEGER | Contains topic identifiers from the model table. |
| topicweight | DOUBLE PRECISION | Contains topic weights. |
| topicwords | TEXT, VARCHAR, or VARCHAR(n) | Optional. Contains topic words, separated by commas. |
Because the the model table contents are in BYTEA format, it is not readable. To see the binary contents, use the function LDATopicPrinter.