BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Timothy Miller

Timothy Miller contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
astro-ph.GAArtificial IntelligenceComputation and Languagecs.CYMachine Learning

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
7works
0followers
5topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

7 published item(s)

preprint2022arXiv

Hierarchical Annotation for Building A Suite of Clinical Natural Language Processing Tasks: Progress Note Understanding

Applying methods in natural language processing on electronic health records (EHR) data is a growing field. Existing corpus and annotation focus on modeling textual features and relation prediction. However, there is a paucity of annotated corpus built to model clinical diagnostic thinking, a process involving text understanding, domain knowledge abstraction and reasoning. This work introduces a hierarchical annotation schema with three stages to address clinical text understanding, clinical reasoning, and summarization. We created an annotated corpus based on an extensive collection of publicly available daily progress notes, a type of EHR documentation that is collected in time series in a problem-oriented format. The conventional format for a progress note follows a Subjective, Objective, Assessment and Plan heading (SOAP). We also define a new suite of tasks, Progress Note Understanding, with three tasks utilizing the three annotation stages. The novel suite of tasks was designed to train and evaluate future NLP models for clinical text understanding, clinical knowledge representation, inference, and summarization.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Summarizing Patients Problems from Hospital Progress Notes Using Pre-trained Sequence-to-Sequence Models

Automatically summarizing patients' main problems from daily progress notes using natural language processing methods helps to battle against information and cognitive overload in hospital settings and potentially assists providers with computerized diagnostic decision support. Problem list summarization requires a model to understand, abstract, and generate clinical documentation. In this work, we propose a new NLP task that aims to generate a list of problems in a patient's daily care plan using input from the provider's progress notes during hospitalization. We investigate the performance of T5 and BART, two state-of-the-art seq2seq transformer architectures, in solving this problem. We provide a corpus built on top of progress notes from publicly available electronic health record progress notes in the Medical Information Mart for Intensive Care (MIMIC)-III. T5 and BART are trained on general domain text, and we experiment with a data augmentation method and a domain adaptation pre-training method to increase exposure to medical vocabulary and knowledge. Evaluation methods include ROUGE, BERTScore, cosine similarity on sentence embedding, and F-score on medical concepts. Results show that T5 with domain adaptive pre-training achieves significant performance gains compared to a rule-based system and general domain pre-trained language models, indicating a promising direction for tackling the problem summarization task.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Deep Representation Learning of Patient Data from Electronic Health Records (EHR): A Systematic Review

Patient representation learning refers to learning a dense mathematical representation of a patient that encodes meaningful information from Electronic Health Records (EHRs). This is generally performed using advanced deep learning methods. This study presents a systematic review of this field and provides both qualitative and quantitative analyses from a methodological perspective. We identified studies developing patient representations from EHRs with deep learning methods from MEDLINE, EMBASE, Scopus, the Association for Computing Machinery (ACM) Digital Library, and Institute of Electrical and Electronics Engineers (IEEE) Xplore Digital Library. After screening 363 articles, 49 papers were included for a comprehensive data collection. We noticed a typical workflow starting with feeding raw data, applying deep learning models, and ending with clinical outcome predictions as evaluations of the learned representations. Specifically, learning representations from structured EHR data was dominant (37 out of 49 studies). Recurrent Neural Networks were widely applied as the deep learning architecture (LSTM: 13 studies, GRU: 11 studies). Disease prediction was the most common application and evaluation (31 studies). Benchmark datasets were mostly unavailable (28 studies) due to privacy concerns of EHR data, and code availability was assured in 20 studies. We show the importance and feasibility of learning comprehensive representations of patient EHR data through a systematic review. Advances in patient representation learning techniques will be essential for powering patient-level EHR analyses. Future work will still be devoted to leveraging the richness and potential of available EHR data. Knowledge distillation and advanced learning techniques will be exploited to assist the capability of learning patient representation further.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

HST/COS Observations of Quasar Outflows in the 500 -- 1050 Å Rest Frame: II The Most Energetic Quasar Outflow Measured to Date

We present a study of the BAL outflows seen in quasar SDSS J1042+1646 (z = 0.978) in the rest-frame 500 -- 1050 $Å$ (EUV500) region. The results are based on the analysis of recent Hubble Space Telescope/Cosmic Origins Spectrograph observations. Five outflow systems are identified, where in total they include $\sim$70 outflow troughs from ionic transitions. These include the first non-solar detections from transitions of O V*, Ne V*, Ar VI, Ca VI, Ca VII, and Ca VIII. The appearance of very high-ionization species (e.g., Ne VIII, Na IX, and Mg X) in all outflows necessitates at least two-ionization phases for the observed outflows. We develop an interactive Synthetic Spectral Simulation method to fit the multitude of observed troughs. Detections of density sensitive troughs (e.g., S IV* $λ$ 657.32 $Å$ and the O V* multiplet) allow us to determine the distance of the outflows ($R$) as well as their energetics. Two of the outflows are at $R$ $\simeq$ 800 pc and one is at $R$ $\simeq$ 15 pc. One of the outflows has the highest kinetic luminosity on record ($\dot{E_{k}}$ $ = 5\times 10^{46}$ erg s$^{-1}$), which is 20% of its Eddington luminosity. Such a large ratio suggests that this outflow can provide the energy needed for active galactic nucleus feedback mechanisms.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

HST/COS Observations of Quasar Outflows in the 500 -- 1050 Å Rest Frame: IV. The Largest Broad Absorption Line Acceleration

We present an analysis of the broad absorption line (BAL) velocity shift that appeared in one of the outflow systems in quasar SDSS J1042+1646. Observations were taken by the Hubble Space Telescope/Cosmic Origin Spectrograph in 2011 and 2017 in the 500 -- 1050 $Å$ rest frame. The outflow&#39;s velocity centroid shifted by $\sim$ --1550 km s$^{-1}$ from --19,500 km s$^{-1}$ to --21,050 km s$^{-1}$ over a rest-frame time of 3.2 yr. The velocity shift signatures are most apparent in the absorption features from the Ne VIII $λλ$ 770.41, 780.32 doublet and are supported by the absorption troughs from OV $λ$ 629.73 and the Mg X $λλ$ 609.79, 624.94 doublet. This is the first time where a quasar outflow velocity shift is observed in troughs from more than one ion and in distinct troughs from a doublet transition (Ne VIII). We attribute the velocity shift to an acceleration of an existing outflow as we are able to exclude photoionization changes and motion of material into and out of the line of sight as alternate explanations. This leads to an average acceleration of 480 km s$^{-1}$ yr$^{-1}$ (1.52 cm s$^{-2}$) in the quasar rest frame. Both the acceleration and the absolute velocity shift are the largest reported for a quasar outflow to date. Based on the absorption troughs of the O V* multiplet, we derive a range for the distance of the outflow ($R$) from the central source, 0.05 pc $<$ $R$ $<$ 54.3 pc. This outflow shows similarities with the fast X-ray outflow detected in quasar PG 1211+143. We use the acceleration and velocity shift to constrain radiatively accelerated active galactic nucleus disk-wind models and use them to make predictions for future observations.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

HST/COS Observations of Quasar Outflows in the 500 -- 1050 Å Rest Frame: VI Wide, Energetic Outflows in SDSS J0755+2306

We present the analysis of two outflows (S1 at --5500 km s$^{-1}$ and S2 at --9700 km s$^{-1}$) seen in recent HST/COS observations of quasar SDSS J0755+2306 (z = 0.854). The outflows are detected as absorption troughs from both high-ionization species, including N III, O III, and S IV, and very high-ionization species, including Ar VIII, Ne VIII, and Na IX. The derived photoionization solutions show that each outflow requires a two ionization-phase solution. For S1, troughs from S IV* and S IV allow us to derive an electron number density, $n_{e}$ = 1.8$\times$10$^4$ cm$^{-3}$, and its distance from the central source of $R$ = 270 pc. For S2, troughs from O III* and O III yield $n_{e}$ = 1.2$\times$10$^3$ cm$^{-3}$ and $R$ = 1600 pc. The kinetic luminosity of S2 is $>$ 12% of the Eddington luminosity for the quasar and therefore can provide strong AGN feedback effects. Comparison of absorption troughs from O III and O VI in both outflow systems supports the idea that for a given element, higher ionization ions have larger covering fractions than lower ionization ones.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Evidence that Emission and Absorption Outflows in Quasars Are Related

We analyze VLT/X-shooter data for 7 quasars, where we study the relationships between their broad absorption line (BAL) and emission line outflows. We find: 1) the luminosity of the [OIII] $λ$5007 emission profile decreases with increasing electron number density (n$_e$) derived from the BAL outflow in the same quasar, 2) the measured velocity widths from the [OIII] emission features and CIV absorption troughs in the same object are similar, and 3) the mean radial velocity derived from the BAL outflow is moderately larger than the one from the [OIII] emission outflow. These findings can be explained by the physical interpretation that the [OIII] and BAL outflow are different manifestations of the same wind. When we have outflows with smaller distances to the central source, their n$_e$ is higher. Therefore, the [OIII] emission is collisionally de-excited and the [OIII] luminosity is then suppressed. Comparisons to previous studies show that the objects in our sample exhibit broad [OIII] emission features similar to the ones in extremely red quasars (ERQs). This might imply that BAL quasars and ERQs have the same geometry of outflows or are at a similar evolutionary stage. We found that the physical parameters derived from the BAL outflows can explain the amount of observed [OIII] luminosity, which strengthens our claim of both BAL and [OIII] outflows are from the same wind. These estimates can be tested with upcoming James Webb Space Telescope observations.

0 citations0 reviewsNo code linkedOpen access