From a file to a list of elements

Elements

Every file type

Cleaning up the text

Chunking

Learn the open source Unstructured library from scratch: partition any file into elements, read their metadata, clean and chunk them, and build a searchable document handbook. Every lesson runs with no API key.

Unstructured

The four things unstructured does

Unstructured tutorial: parse documents into elements in Python

What you are going to build

Parsing a document without Unstructured, and what it costs

Reading a document with plain Python

Why hand-written parsers break, and what Unstructured does

Where hand-written parsing breaks

Install Unstructured and partition your first document

Install unstructured, and partition one file

Inspect Unstructured elements: category, text and to_dict

Looking at what came back

How one line of plain text gets its label

Unstructured element types: Title, NarrativeText, Table

Element types, and what decides them

Unstructured metadata: filename, filetype, languages and more

Metadata: where an element came from

Rebuild a document outline with Unstructured parent_id

parent_id: the document's own outline

Read tables with Unstructured: text_as_html and what it loses

Tables, and text_as_html

Unstructured element_id: what it hashes and when it changes

Element ids, and what changes them

How a file's type is decided

How Unstructured detects a file type and picks a partitioner

How partition picks a partitioner

Unstructured extras: which file types your install reads

What this install can read, and what it cannot

Unstructured PDF strategies: fast, hi_res and ocr_only

PDF and images: the extra this course skips

Partition Word, PowerPoint and email files with Unstructured

Email, Word and PowerPoint

Unstructured partition from a file object, and text=

Partitioning a stream instead of a path

Clean Unstructured element text with cleaners

Cleaners, and element.apply

Pull emails, phones and dates from Unstructured elements

Pulling values out of text

Filter Unstructured elements by type with filter_element_types

Dropping elements you do not want

Unstructured chunking: chunk_elements and CompositeElement

Why elements are not chunks

Unstructured chunk_by_title and combine_text_under_n_chars

chunk_by_title: splitting where the document splits

Unstructured chunk overlap, orig_elements and table chunks

Overlap, combining, and orig_elements

From a file to a list of elements quiz

The lesson 1 splitter found four blocks in <code>shipping.md</code>. <code>partition</code> found eight elements in the same file. Why?