# U.S. Senate Global Media Intelligence and Retrieval System ## One-sentence description At the Triad Center in Salt Lake City, I helped architect and integrate Senate-facing systems that converted live international television and governmental information into continuously ingested, time-associated, machine-searchable institutional memory while the public Web and real-time information retrieval were still emerging. ## The project and agreement During the Triad Center period, John “Fred” Adamson paid me to develop the technology for a Senate-facing information system. The project had an initial greenlight, and the first demonstrations were received enthusiastically. Successful delivery was expected to lead to the government contract and ownership shares for me. The venture later operated through **Intelenet Inc.**, incorporated in April 1995. A June 8, 1996 *Deseret News* report, [“Intelenet's Demise Sinks Surfers”](https://www.deseret.com/1996/6/8/19247470/intelenet-s-demise-sinks-surfers/), describes Intelenet as a local Internet service company providing dial-up access and Web-page service from the Triad Center. It ceased active business in early June 1996. ## Problem it addressed A live television broadcast is an ephemeral linear event. A person must watch it when it happens, record it for later review, or lose access to what was said. The problem becomes qualitatively different when the source field is not a few domestic channels but hundreds of international broadcasts. The Senate-facing media system addressed that problem by converting broadcast captions into searchable information. Instead of requiring analysts to watch many simultaneous feeds or manually review videotape, the system extracted text already embedded in the broadcast, associated it with source and time context, processed it, and made it rapidly retrievable. A related problem involved governmental information that lived in documents, records, correspondence systems, files, and separate repositories. Senate users needed information to become computationally addressable through search and emerging network interfaces. ## Human purpose and users The human purpose was to reduce the distance between an institutional question and the information needed to answer it. A Senate user interested in a country, person, organization, phrase, legislative subject, or breaking event could search a growing textual record rather than locate and inspect hours of recorded programming. The document and correspondence work served a parallel purpose: making legislative and governmental records easier to find, retrieve, disseminate, and potentially expose through emerging digital public-information systems. ## What I helped build The most technically specific current description is a real-time global media-ingestion and retrieval environment that: - monitored more than 200 international television channels through satellite or broadcast feeds; - extracted closed-caption text; - normalized or processed the resulting information; - supported multilingual processing; - associated content with its originating source and temporal context; - continuously indexed the textual stream; - and enabled sub-second or extremely low-latency search for legislative and media intelligence. Parallel Senate-facing initiatives involved search and retrieval for legislative documents, records, and correspondence and broader work around networked access to governmental information. I engineered and integrated major portions of this environment within a team spanning media reception, telecommunications, search, data processing, and institutional relationships. ## Language and retrieval toolchain My principal implementation and integration language was **Perl**, running on a **custom build of Red Hat Linux**. It handled ingestion, transformation, indexing, text processing, intermediate representations, command-line coordination, and the integration of systems that had never been designed as one stack. The surrounding full-text and document-retrieval environment included: - **dtSearch** for Windows and local-network full-text retrieval; - **Glimpse** for indexed search across large Unix file collections; - **Folio Views / Folio Infobase** for structured, distributable full-text corpora with hierarchy and hypertext links; - and **Symantec Q&A** for databases, forms, networking, substantial text fields, and English-like retrieval through its Intelligent Assistant. Primary historical references for the tool descriptions include [dtSearch's official history](https://www.dtsearch.com/dtsoftware.html), the University of Arizona's [1993 Glimpse technical report](https://www.cs.arizona.edu/sites/default/files/TR93-34.pdf), the [1994 USENIX Glimpse record](https://www.usenix.org/conference/usenix-winter-1994-technical-conference/glimpse-tool-search-through-entire-file-systems), Cornell's 1994 account of [moving Folio Views infobases to HTML](https://www.law.cornell.edu/papers/lii/fffhtml.htm), and a preserved 1992 review of [Q&A 4.0 and its Intelligent Assistant](https://www.aallnet.org/litsis/wp-content/uploads/sites/5/2018/05/CS-SIS-AUTOMATOME-11.1-WINTER-1992.pdf). The room itself was heterogeneous. My custom Red Hat and Perl environment supplied the core server-side and integration layer, while Windows and DOS workstations, modem-connected users, Unix and NeXT systems, structured document collections, and emerging Web interfaces required different retrieval geometries. We were searching arbitrary file collections, structured infobases, application databases, and increasingly network-accessible information spaces across mixed operating systems. ## 1996 continuation after the Triad Center I experimented with **WebGlimpse in 1996**, after the Triad Center environment had fallen apart. That year, Trevor Gillette Smith lived with me in Spring City, Utah. We continued working together with **Linux and Perl**, carrying the Triad technical relationship into a smaller and independent setting. The connection is still historically meaningful. WebGlimpse extended Glimpse-style indexed retrieval into web-accessible and hierarchically related collections, matching the direction in which the earlier document and file-search work had been moving. A [1996 W3C workshop paper](https://www.w3.org/Search/9605-Indexing-Workshop/Papers/Manber%40Arizona.html) describes WebGlimpse as combining browsing and searching through site-level indexes and page neighborhoods. The work became a personal technical continuation of the retrieval lineage after Triad. ## Information pipeline The media pipeline moved through these stages: **live television and satellite sources → broadcast reception → closed-caption extraction → text normalization and language processing → source/time association → continuous indexing → query and retrieval** ## The second information pipeline A second pipeline handled legislative documents, records, and correspondence: **legislative documents, records, and correspondence → ingestion or conversion → normalization and metadata → indexing → network search and retrieval → internal or public information access** Both activities transformed isolated or transient information into searchable institutional memory. ## The Triad Center environment This work did not occur inside a modern cloud abstraction. I remember a hardware-dense communications environment in which telecommunications, broadcast engineering, computing, and networking met physically. The room-level recollection includes: - satellite dishes and incoming live television feeds; - NeXT computers or workstations; - Silicon Graphics UNIX workstations; - a large modem plant remembered at approximately 300 devices, much of it Hayes-era equipment; - early Internet or upstream network connectivity; - mixed computer and communications hardware that had to be made interoperable; - a custom Red Hat Linux build used as the core runtime for my Perl-based work, whose installation and configuration included an effort remembered as lasting roughly three sleepless days; - work in search, networking, ingestion, information processing, and data compression. ## Multilingual processing Multilingual processing made international material addressable inside the wider retrieval environment. It joined language handling to source identity, time context, and continuous indexing. ## Latency and scale Two figures convey the project's scale and speed: - **more than 200 international television channels**; - **sub-second indexing or search**. Together they describe an environment designed to turn a vast live source field into information that could be searched while it was still operationally useful. ## Human team Trevor Gillette Smith collaborated with me on Senate data ingestion. In 1996, after the Triad operation had fallen apart, Trevor lived with me in Spring City, Utah, and we continued working together with Linux and Perl. **John “Fred” Adamson** originated the Senate project and handled the connections and institutional relationships while I handled the technology. His family called him Fred. John and Trevor later died together as a result of the same automobile crash in Mill Creek Canyon. Trevor died from his injuries on January 18, 2001. Trevor had met John in Salt Lake City and worked at the Electronic Cottage on and off for years. ## Modern architectural translation For a contemporary technical reader, the closest analogies are: - streaming-data ingestion; - event-processing pipelines; - media monitoring and observability; - open-source intelligence platforms; - social-listening systems; - enterprise search; - time-indexed corpora; - and retrieval-augmented institutional intelligence. A modern pipeline often looks like: **source → ingest → normalize → enrich → timestamp → index → query → alert or analyze** The Triad environment embodied a recognizable earlier version using satellite television rather than social APIs, closed captions rather than transformer-based speech transcription, UNIX workstations rather than cloud containers, modem banks rather than ubiquitous broadband, and custom retrieval infrastructure rather than Elasticsearch or vector databases. Today I recognize the same functional pattern in streaming data, media monitoring, enterprise search, social listening, OSINT, and retrieval-augmented institutional intelligence. ## What the system accomplished conceptually The central transformation was simple to state and difficult to engineer: **broadcast became data; data became indexed text; indexed text became searchable institutional memory.** At the same time, governmental documents and correspondence were undergoing a related transformation from isolated artifacts into network-retrievable information objects. The unifying project was therefore not television by itself and not search by itself. It was the creation of a continuously ingestible, searchable information space from heterogeneous streams and repositories. ## Relationship to later work This project is an early anchor for several patterns that continued through my later systems: - transient information converted into durable addressable form; - multiple sources normalized into a common retrieval surface; - expensive processing performed before the user asks a question; - source identity and temporal context preserved alongside text; - heterogeneous hardware and software joined through explicit interfaces; - and information systems designed around the human time saved by retrieval. Those same patterns recur in the [[projects/Ten Years Building a Symbolic Language Engine|Symbolic Language Engine]], [[projects/Shazam - A Portable Perl Capability Library|Shazam]], WordChirp, [[projects/AutoSites - Folder-Native Publishing and Distribution|AutoSites]], and the present Obsidian knowledge system.