What DOI, title, and abstract mean in OpenAlex

OpenAlex is an open index of scholarly works. A “work” is usually a journal article, conference paper, preprint, or book chapter. The OpenAlex API documents works as metadata records: who wrote it, where it appeared, what it is called, and, when the source allows it, a short summary. conscRAG does not host that library. When you find a work or run a title, we query OpenAlex and read three fields first: title, DOI, and abstract.

Those three fields are how most people recognize a paper. They are also how retrieval works. A search engine can match words in a title. A DOI is a stable pointer. An abstract is the only full-text-like passage many indexes are allowed to store. If you have used Crossref, PubMed, or a university library catalog, you have already seen the same trio under slightly different names.

In OpenAlex the title is stored as display_name. It is the human-readable name of the work, not a filename and not a citation key. Two records can share a similar title (think of every paper that riffs on “attention is all you need”). The title is still the best first handle: it is what you type into Find a paper, and it is what conscRAG sends back into OpenAlex when it retrieves related work.

A title is not a unique identifier. Translations, preprints, and later journal versions can differ by a subtitle or a word. That is why OpenAlex also keeps ids (an OpenAlex id, and often a DOI). When you click a hit on conscRAG we copy the display_name into the Title field so Run starts from the same string the index used.

The International DOI Foundation puts it plainly: a DOI is “a persistent identifier for a digital object.” It is a string that starts with 10., such as 10.48550/arxiv.2305.14259. Registration agencies like Crossref and DataCite mint them. The resolver at doi.org turns that string into the publisher or repository page. OpenAlex stores the DOI when it knows one. Many older preprints, theses, and some books never received one.

conscRAG treats a DOI as a lookup key, not as a download. Find a paper can resolve doi:… or a doi.org URL. On Run, if you leave title or abstract blank but supply a DOI, we call OpenAlex works/doi:… and fill the missing text. If OpenAlex has the DOI but no usable abstract, Run stops: we will not invent a summary. See how we process a run.

An abstract is the author’s short account of the problem, method, and result. OpenAlex often stores it as an inverted index (each word mapped to positions) and rebuilds the sentence order. conscRAG reconstructs that text, then truncates it for retrieval and for the form. Generate still requires a real abstract of meaningful length. Find a paper will list a hit even if the abstract is missing, and mark “no abstract” so you can paste one yourself.

Not every OpenAlex record has an abstract. Publisher licenses, incomplete deposits, and books without a stand-alone summary all leave gaps. That is a coverage limit of the index, not a bug in the form. If the abstract is missing, the honest move is to paste it from the PDF or the publisher page you are allowed to use.

Title without abstract is a label. Abstract without title is a paragraph you cannot cite. DOI without either is a pointer you still have to open. conscRAG’s drop form keeps all three because the pipeline needs a title to search and an abstract to profile. The DOI is optional glue: it lets you skip typing, and it lets someone else reopen the same work later.

If you want the product-level view of identifiers beyond DOI, read what title, keywords, DOI, arXiv id, and PMID mean in Find a paper. OpenAlex’s own API docs are at docs.openalex.org.